早期算法

EARLY ALGORITHMS

算法，作为一个术语，简单地描述了一个方程式：任何能够产生预期结果的公式或规则集。最早的例子来自古巴比伦，位于现在的伊拉克地区。楔形文字泥板可以追溯到公元前1800年至1600年，记录了一些算法，例如利用蓄水池的深度和为其挖出的土方量来计算其长度和宽度。根据数学家唐纳德·E·克努斯（Donald E. Knuth）的说法，巴比伦人“用一系列逐步的规则来表示每个公式，以便对其进行评估，即用一个计算该公式的算法”。他们有一个专门的计算记录系统，使用“公式的‘机器语言’表示，而不是符号语言”，克努斯写道。每个巴比伦算法的书面解释都以同一句话结尾：“这就是程序”。这句话强调了算法的内在特性：它们可以重复使用，在每次特定情况下都同样适用且有效。如今硅谷的追随者可能会将它们描述为可扩展的。

Algorithm as a term simply describes an equation: any formula or set of rules that produces a desired result. The earliest examples come from ancient Babylon, in the region that is now Iraq. Cuneiform tablets, dating back to 1800–1600 BCE, record algorithms for purposes like calculating the length and width of a cistern using its depth and the volume of earth excavated for it. According to the mathematician Donald E. Knuth, the Babylonians “represented each formula by a step-by-step list of rules for its evaluation, i.e., by an algorithm for computing that formula.” They had a specialized system for recording calculations, using “a ‘machine language’ representation of formulas instead of a symbolic language,” Knuth wrote. The written explanation of each Babylonian algorithm ended with the same phrase: “This is the procedure.” That line emphasizes an inherent quality of algorithms: they can be repeated, equally applicable and effective every time a given situation occurs. An acolyte of Silicon Valley today might describe them as scalable.

算法是早期数学史的关键。大约公元前300年，希腊哲学家欧几里得在他的著作《几何原本》中记录了所谓的欧几里得算法，这是一种寻找两个或多个数的最大公约数的方法。该公式和埃拉托斯特尼筛法（一种公元前三世纪用于识别一组数中的素数的算法）是至今仍在使用，尤其是在密码学领域。但“算法”这个词实际上来自一个人——或者至少是他的出生地。

Algorithms are key to the history of early mathematics. Around 300 BCE, the Greek philosopher Euclid recorded in his treatise Elements what is called the Euclidean algorithm, a way of finding the greatest common divisor of two or more numbers. That formula and the Sieve of Eratosthenes, an algorithm from the third century BCE that identifies prime numbers within a set of numbers, are still used today, particularly in the realm of cryptography. But the actual word algorithm comes from a single person—or at least his birthplace.

穆罕默德·伊本·穆萨·花拉子密是一位波斯学者，约公元 780 年出生于花剌子模，该地区位于今土库曼斯坦和乌兹别克斯坦附近。虽然人们对他的生平知之甚少，但花拉子密曾前往巴格达。在七世纪穆斯林阿巴斯王朝征服波斯后，巴格达成为该地区的知识中心。他在智慧宫（又称巴格达大图书馆）工作，研究占星术、地理和数学。与其前身埃及亚历山大图书馆一样，智慧宫是一个跨学科的学习中心，重视科学研究，并将希腊语、拉丁语、梵语和波斯语的文献翻译成阿拉伯语。大约公元 820 年，花拉子密完成了《印度数字计算法》，这部著作最终将我们今天使用的数字系统引入了欧洲。他还撰写了《复原与化简规则》，这是一本关于方程解法的书。它的阿拉伯语名称被缩写为al-jabr（意为“复原”，或取消等式两边的同类项），这为“复原与化简”一词和代数学科的诞生提供了源头。《复原与化简规则》包含二次方程的解法以及面积和体积的计算方法，并给出了π的近似值。

Muhammad ibn Musa al-Khwarizmi was a Persian scholar born around 780 CE in Khwarazm, an area around present-day Turkmenistan and Uzbekistan. Though little is known about his life, Al-Khwarizmi made his way to Baghdad, which had become the intellectual center of the region after the Muslim Abassid caliphate conquered Persia in the seventh century. There he worked at the House of Wisdom, also known as the Grand Library of Baghdad, researching astrology, geography, and math. Like its predecessor the Egyptian Library of Alexandria, the House of Wisdom was an interdisciplinary center of learning where scientific study was prized and texts in Greek, Latin, Sanskrit, and Persian were translated into Arabic. Around 820, al-Khwarizmi completed On the Calculation with Hindu Numerals, the text that eventually introduced the numeral system we use today to Europe. He also wrote The Rules of Restoration and Reduction, a book on strategies for solving equations. Its Arabic name was shortened to al-jabr (meaning restoration, or canceling like terms on either side of an equation), which provided the source for the word and the discipline of algebra. Restoration and Reduction included solutions for quadratic equations and methods of calculating area and volume, with approximations of pi.

十二世纪中叶，一位名叫罗伯特·切斯特的英国阿拉伯语学者旅居西班牙。当时，穆斯林、犹太教和基督教文化在此交融，时而和平共处，时而又不那么和谐。这又是一个思想在不同文明之间交流传播的时期。1145年，罗伯特将《复原与还原规则》译成拉丁语。“Al-jabr”变成了“algeber”（代数师），而“al-Khwarizmi”变成了“Algoritmi”（算法师）。当时，“algorismus”（算法学）泛指任何使用印度-阿拉伯数字的数学程序，从事此类工作的人被称为算法学家。（该术语最初由使用算法过程的视觉艺术家于20世纪60年代开始采用，但似乎也适用于任何研究当今算法的人。）算法的悠久词源表明，计算既是人类艺术和劳动的产物，也是可重复的科学定律的产物。

In the mid-twelfth century, an English scholar of Arabic named Robert of Chester was living in Spain, where Muslim, Jewish, and Christian cultures overlapped, at some times peacefully and at others less so. It was another moment when ideas were exchanged and disseminated, crossing between civilizations. In 1145, Robert translated The Rules of Restoration and Reduction into Latin. Al-jabr became “algeber,” and al-Khwarizmi became “Algoritmi.” At that time, “algorismus” referred generally to any kind of mathematical procedure using Hindu-Arabic numerals, and those who practiced such an art were called algorists. (That term was adopted by visual artists using algorithmic processes beginning in the 1960s, but it seems apt for anyone working on today’s version of algorithms.) The long arc of algorithm’s etymology shows that calculations are a product of human art and labor as much as repeatable scientific law.

计算机编程的发明

THE INVENTION OF COMPUTER PROGRAMMING

所有计算机都是由一系列反复执行的方程式构建而成。结果以零和一编码，然后通过更多方程式传递以得出结果。1822年，英国发明家查尔斯·巴贝奇概述了他的“将机械应用于天文和数学表格计算”的概念——一种使用一组编号轮子和齿轮（称为差分机）来自动执行计算的方法。这台机器从未完全制造出来，但后来的执行过程看起来有点像钢琴的内部，只是用一排排的轮子代替了锤子。巴贝奇的设计将有八英尺高，四吨重。他后来的迭代产品——分析机——如果能够制造出来，可以接受通过穿孔卡片编写的命令，并执行循环和条件等简单的编程功能。它为之后所有更为复杂的现代计算奠定了基础。正如巴贝奇的儿子亨利在1888年所写：“这只是卡片和时间的问题。”

All computers are built from series of equations performed repeatedly. Results are encoded in zeros and ones and then passed on through yet more equations to achieve an outcome. In 1822, the British inventor Charles Babbage outlined his concept for the “application of machinery to the computation of astronomical and mathematical tables”—a way to automate calculations using an assemblage of numbered wheels and gears called the Difference Engine. The machine was never fully built, but later executions look something like the inside of a piano, with wheels in long rows instead of hammers. Babbage’s design would have been eight feet high and weighed four tons. His later iteration, the Analytical Engine, could, if built, take commands programmed via punch cards and perform simple programming features like loops and conditions. It was the basis for all the much more complicated modern computing that followed. As Babbage’s son Henry wrote in 1888, “It is only a question of cards and time.”

拜伦勋爵的女儿艾达·洛夫莱斯如今被广泛认为是第一位计算机程序员；她为巴贝奇设计的这台机器编写了算法，其中包括计算伯努利数的过程。洛夫莱斯还意识到，这台机器所实现的重复机械过程可以应用于数学以外的领域。1843年，洛夫莱斯写道，分析机“除了数字之外，还可以作用于其他事物，只要这些事物的相互基本关系可以用抽象的运算科学来表达，并且也应该能够适应分析机的操作符号和机制。” 换句话说，任何可以转化为数据（一系列数字）的东西，都可以用公式化的方式进行操作。这可能包括文本、音乐、艺术，甚至像国际象棋这样的游戏。洛夫莱斯设想了一种这样的自动化形式：“例如，假设和声学和音乐创作中音调的基本关系能够被如此表达和改编，那么引擎就可以创作出任何复杂程度或范围的精致而科学的音乐作品。”她设想类似于作曲家布莱恩·伊诺在1995年创作并推广的“生成音乐”，即一系列由音乐软件驱动的氛围合成器作品，每次运行时都会产生不同的旋律。洛夫莱斯设想的是，文化如何被新技术塑造和延续，就像今天的算法信息流一样。

Ada Lovelace, the daughter of Lord Byron, is now widely regarded as the first computer programmer; she wrote algorithms for the machine as Babbage designed it, including a process for calculating Bernoulli numbers. Lovelace also realized that the repeating mechanical processes that the machine enabled could be applied to fields beyond mathematics. In 1843, Lovelace wrote that the Analytical Engine “might act upon other things besides number, were objects found whose mutual fundamental relations could be expressed by those of the abstract science of operations, and which should be also susceptible of adaptations to the action of the operating notation and mechanism of the engine.” In other words, anything that can be turned into something like data—a series of numbers—could be manipulated in a formulaic way. That might include text, music, art, or even a game like chess. Lovelace imagined one form of such automation: “Supposing, for instance, that the fundamental relations of pitched sounds in the science of harmony and of musical composition were susceptible of such expression and adaptations, the engine might compose elaborate and scientific pieces of music of any degree of complexity or extent.” She was envisioning something like what the composer Brian Eno created in 1995 and popularized as “generative music,” a series of ambient synth compositions driven by musical software that created different tunes each time the software ran. Lovelace was envisioning how culture could be both molded and perpetuated by the new technology, the way algorithmic feeds do today.

洛夫莱斯很早就发现，操控这些机械指令可以成为其自身的形式和自我表达。在20世纪90年代和21世纪初，计算机编程开始与基础数学和科学并驾齐驱，成为儿童完整教育的必备技能。大约在2002年，我在高中“计算机室”的电脑上接触到了计算机编程，当时我们玩一些类似编程语言的教育类电子游戏。但我真正学习编程的，是那些笨重的塑料TI-83计算器，那是我们上高等数学课必须买的。这些计算器能够使用一种名为TI-BASIC的语言进行编程，这种语言包含简单的if-then循环和变量函数。起初，我编写了一些简单的程序来自动执行考试所需的公式，但当我对这门语言更加熟练后，我开始制作自己的井字棋和四子棋游戏。这台机器是我创造力的伙伴；感觉就像魔法一样。

Lovelace was early in discovering that manipulating such mechanical commands could be its own form and self-expression. In the 1990s and 2000s, computer programming began to take its place alongside basic math and science as a skill that was necessary for a child’s complete education. I was introduced to it on “computer room” desktops in my high school circa 2002, where we played educational video games that resembled programming languages. But where I really learned was on the chunky plastic TI-83 calculators that we had to acquire for advanced math classes. The calculators came with the ability to code in a language called TI-BASIC, which had simple if-then loops and variable functions. At first, I made modest programs to automate formulas I needed to know for tests, but once I became more fluent in the language, I made my own versions of tic-tac-toe and Connect Four. The machine was a partner in my creativity; it felt like magic.

洛夫莱斯事件发生一个世纪后，二战期间，英国数学家兼计算机科学家艾伦·图灵为政府从事密码破译工作——他参与破译了德国恩尼格玛密码机。1946年，战争结束后，图灵为美国国家物理图书馆撰写了一份报告，提议开发“自动计算机器”。这是人工智能首次被描述为一种真实的可能性，而非一个理论概念。图灵写道，用于执行特定任务的计算和分类机器早已存在，但他的提议远不止于此：“无需重复人工从机器中取出材料并在适当的时候放回，所有这些都将由机器自行完成。”

A century after Lovelace, during World War II, Alan Turing, a British mathematician and computer scientist, was working in code breaking for the government—he helped to decode the German Enigma cipher machine. In 1946, with the war over, Turing wrote a report for the National Physical Library proposing the development of an “Automatic Computing Engine.” It was the first description of artificial intelligence as a real possibility instead of a theoretical concept. Calculating and sorting machines designed to perform specific tasks already existed, Turing wrote, but his proposal went beyond that: “Instead of repeatedly using human labor for taking material out of the machine and putting it back at the appropriate moment all this will be looked after by the machine itself.”

根据图灵的说法，该设备将能够执行任何类型的计算，并且规模不限，无需重新配置。它拥有自己的内部逻辑语言，可以适应不同的目的，解决任何类型的问题。“人们怎么能指望“一台机器能做这么多种类繁多的事情吗？”图灵写道。“答案是，我们应该认为机器正在做一些非常简单的事情，即执行以它能够理解的标准形式给出的命令。”它会执行算法。他暗示了当今机器学习算法随着时间的推移而发展的方式，无需人类决策即可进行调整。

According to Turing, the device would be able to perform any kind of calculation and at any scale without needing to be reconfigured. It had its own internal logical language that could be adapted to different ends, to solve any type of problem. “How can one expect a machine to do all this multitudinous variety of things?” Turing wrote. “The answer is that we should consider the machine to be doing something quite simple, namely carrying out orders given to it in a standard form which it is able to understand.” It would execute algorithms. He hinted at the way that machine learning algorithms today evolve over time, incorporating adjustments without human decision-making.

这样的系统将能够以更快的速度和更高的复杂度执行计算，超越人类。图灵写道：“机器的速度不再受人类操作员速度的限制。” 然而，他并不认为这样的机器是乌托邦式的工具。自动化并不意味着它们总是正确的。图灵继续说道：“人为的易错性被消除了，尽管它可能在某种程度上被机械的易错性所取代。” 他的报告预测了个人计算机的许多如今为人熟知的元素，从可擦除存储单元到输入机制、二进制语言的转换，甚至温度控制（以防止机器过热）。然而，对图灵来说，“计算机”一词指的并非机器，而是指进行计算的人，这再次强调了计算机的有机性。

Such a system would perform calculations much faster and at higher levels of complexity, exceeding humans. Turing wrote: “The speed of the machine is no longer limited by the speed of the human operator.” He did not see such machines as utopian tools, however. Just because they would be automatic didn’t mean they would always be right. “The human element of fallibility is eliminated, although it may to an extent be replaced by mechanical fallibility,” Turing continued. His report predicted many now-familiar elements of personal computers, from erasable memory units to input mechanisms, conversion from binary language, and even temperature control so the machine wouldn’t overheat. For Turing, however, the word computer referred not to the machine but to the person doing the computing, once again emphasizing that organic element.

早在1936年，图灵就构想出了现在所谓的“图灵机”，并在1948年题为《智能机械》的论文中对其进行了详细描述。图灵机是“一条被划分成方格的无限长纸带，每个方格上都可以印上一个符号”。纸带在读取器中移动，读取器一次扫描一个方格，并执行方格中符号指示的操作，这些符号也可以被擦除或覆盖。任何算法，从数学过程的历史意义上来说，都可以用这样的图灵机来计算。任何能够计算图灵机所能计算的一切的计算系统都被称为“图灵完备的”。例如，所有编程语言都是图灵完备的，因为它们可以模拟任何类型的方程。 （甚至电子表格软件 Excel 也在 2021 年实现了图灵完备。）图灵正确地得出结论：任何计算机器都能够完成任何其他计算机器的工作——即使是查尔斯·巴贝奇 (Charles Babbage) 19 世纪的分析机，如果给予无限的规模和时间，理论上也可以执行我们的笔记本电脑现在所做的复杂任务。

As early as 1936, Turing conceived of what is now called a “Turing machine,” which he sketched in detail in a 1948 essay called “Intelligent Machinery.” The Turing machine is “an infinite tape marked out into squares, on each of which a symbol could be printed.” The tape moves through a reader that scans one square at a time and performs the operation dictated by the symbol in the square, which can also be erased or overwritten. Any algorithm, in the historical sense of a mathematical process, can be calculated by such a Turing Machine. And any computational system that can compute anything that a Turing Machine can is said to be “Turing-complete.” All programming languages, for example, are Turing-complete because they can model any kind of equation. (Even the spreadsheet software Excel became Turing-complete in 2021.) What Turing correctly concluded was that any computing machine would be able to do the work of any other—even Charles Babbage’s nineteenth-century Analytical Engine could theoretically perform the complex tasks that our laptops do now, if given infinite scale and time.

图灵的生活中也存在着机械规则与人性操作之间的冲突。1952年，图灵因“同性恋表达”（法律术语，指与另一名男子发生性关系）被控严重猥亵罪。在他自己的住所被盗后，他发起了一场混乱的法律诉讼。在英国，成年人之间自愿的同性恋行为一直到1967年都是非法的——法律本身就是一套算法，基于一套不可动摇的规则来做出判决。图灵最终对指控认罪并被定罪。他没有被监禁，而是被迫接受化学阉割。1954年6月，41岁的图灵被他的管家发现死亡。死因是氰化物中毒，他的死因长期以来被认为是自杀，疑似自杀的原因是图灵床边放着一个吃了一半的苹果。

There is something of the clash between mechanical rules and human operation within Turing’s life, too. In 1952, Turing was charged with gross indecency for “homosexual expression”—the legalistic phrase for having sex with another man—during messy legal proceedings that he initiated after his own house was robbed. Homosexual sex among consenting adults remained illegal in England all the way to 1967—the law is its own kind of algorithm, deciding judgment based on an implacable set of rules. Turing eventually pled guilty to the charges and was convicted. Rather than be imprisoned, he was forced to undergo chemical castration. In June 1954, Turing, forty-one, was found dead by his housekeeper. The cause was cyanide poisoning, and his death has long been considered suicide, the suspected delivery mechanism a half-eaten apple at Turing’s bedside.

当我们谈论“算法”时，它常常感觉像是最近才在社交网络时代出现的一种力量。但我们讨论的是一项有着悠久历史和遗产的技术，它经过几个世纪的缓慢形成，远早于互联网的出现。重新审视这个更大的图景可以帮助我们更好地理解算法在当今的力量。然而，无论算法多么复杂，它的本质都是一个方程式：一种得出期望结论的方法，无论是将一定数量的谷物平均分配给几个人的苏美尔图表，还是决定在你打开网站时首先显示哪条帖子的 Facebook 信息流。所有算法都是自动化的引擎，正如 Ada Lovelace 所预测的那样，自动化现在已经渗透到我们生活的方方面面，超越了纯数学的范畴。

When we talk about “the algorithm,” it often feels like a force that began to exist only recently, in the era of social networks. But we’re discussing a technology with a history and legacy that has slowly formed over centuries, long before the Internet existed. Restoring this larger picture can help us better understand the power that algorithms have today. Still, no matter how complex, an algorithm remains in its essence an equation: a method to arrive at a desired conclusion, whether it’s a Sumerian diagram to divide an amount of grain equally among several men or the Facebook feed determining which post to show you first when you open the website. All algorithms are engines of automation, and, as Ada Lovelace predicted, automation has now moved into many facets of our lives beyond pure mathematics.

算法决策

ALGORITHMIC DECISION-MAKING

1971年，在智利圣地亚哥市中心的一栋办公楼里，建造了一间六角形的房间，作为全国的控制室。房间的木板墙上装饰着显示器和背光显示屏，显示着国家原材料供应和劳动参与率等指标的数据。房间中央摆放着七把椅子，彼此相对，围成一圈，这些椅子是白色的玻璃纤维翼背座椅，形似船长椅。在一艘科幻太空巡洋舰中。每把椅子的右侧都有一个控制面板，用于操作各个屏幕，还有一个烟灰缸和一个杯架，可能用来放一杯威士忌。这个房间在更大的控制协同工程的支持下被命名为作战室，是由智利社会主义总统萨尔瓦多·阿连德设计的，斯塔福德·比尔担任顾问，比尔是一位英国人，在自己的国家，他因将“控制论”的实践应用于企业管理而闻名。比尔将控制论描述为“控制的科学”。它涉及分析复杂系统，无论是公司还是生物，并确定它们的工作原理，以更好地建模或创建这种智能的自我修正系统。 （在美国，兰德公司早在 20 世纪 50 年代就率先开展了类似的系统分析实践。）“控制协同项目”旨在提供一个理想的模型，实时协助智利政府的决策者们，让他们坐在房间里抽烟、喝威士忌——这又是一场冰冷的技术与混乱的人性的交汇。在房间里，他们观察着那些掌控着国家的算法。

In 1971, in Santiago, Chile, a hexagonal room designed as a kind of control room for the entire country was built in a downtown office building. Monitors and backlit displays adorned the room’s wood-paneled walls, displaying data readouts with metrics like national raw material supplies and labor participation rates. Seven chairs were arrayed in a circle facing each other in the center of the room, white fiberglass wingback seats that resembled the captain’s chair in a science-fictional space cruiser. Each chair had a control panel on the right-hand side to navigate the various screens, as well as an ashtray and a cupholder, perhaps for a tumbler of whiskey. The room, which was named the Operations Room under the aegis of the larger Project Cybersyn, was designed under the socialist Chilean president Salvador Allende and the consultancy of Stafford Beer, a British man who, in his own country, was known for applying the practice of “cybernetics” to business management. Beer described cybernetics as “the science of control.” It involves analyzing complex systems, whether corporations or biology, and determining how they work to better model or create such intelligent, self-correcting systems. (In the United States, a similar practice of systems analysis was pioneered by the RAND Corporation in the 1950s.) Project Cybersyn was meant to provide an ideal model, aiding the Chilean government’s decision-makers in real time as they sat in the room and smoked their cigarettes and drank their whiskey—another meeting of the coldly technological and the messily human. From the room, the men watched the algorithms that oversaw the nation.

由德国顾问桂·邦西普（Gui Bonsiepe）主导的“控制合成计划”（Cyber​​syn Project）的物理设计，营造出一种中世纪现代主义乌托邦主义的氛围。显示器悬浮于墙面，连接显示器和座椅的底层线路隐藏于视野之外。驾驶舱座椅本身造型流畅统一，流畅地弯曲成一体成型的形状。这个房间象征性地将政府职能简化为数据操控，就像赢得一场电子游戏一样。“控制合成计划”承诺用技术监督取代人类领导，寥寥无几的屏幕包含了所有可能需要的信息。你可以坐在其中一张椅子上，观察这个国家发生的一切。

Project Cybersyn’s physical design, led by the German consultant Gui Bonsiepe, created an image of mid-century modernist utopianism. The monitors floated on the walls, the underlying wiring connecting them to the chairs hidden from view. The cockpit chairs themselves were sleek and uniform, smoothly curving in one molded form. The room symbolically reduced government down to the manipulation of data, like winning a video game. Project Cybersyn promised to supplant human leadership with technological oversight, the scant few screens encompassing any information that might be needed. You could sit in one of the chairs and observe everything happening in the country.

然而，赛博协同控制工程的技术只是个幌子，类似于“设计小说”——一种对可能实现的互动幻象。它承诺的功能在当时的计算机网络环境下尚无法实现。它的数据幻灯片是手工创建的，而非自动生成的。它只在一台计算机上运行，​​由智利工厂用来通过电话线发送信息的电传打字机提供数据。最终，虽然房间最终完工，但却从未投入使用。1973年9月11日，在在美国中央情报局的领导下，阿连德政府被推翻，奥古斯托·皮诺切特上台。

However, Project Cybersyn’s technology was a facade, something akin to “design fiction”—an interactive illusion of what might be possible. What it promised was not yet feasible with computer networks at the time. Its data slideshows were created by hand, not automatically generated. It ran on a single computer, fed by telex machines that Chilean factories could use to send information over telephone lines. And finally, though the room was completed, it was never put into action. On September 11, 1973, with the assistance of the United States CIA, Allende’s government was overthrown, and Augusto Pinochet took over.

控制论协同计划（Project Cyber​​syn）的照片依然拥有不可否认的吸引力。它们反复出现在设计情绪板上，投射出一种数十年后依然清晰可见的美感。或许，这些图像如此具有影响力，是因为我们依然怀揣着这样的梦想：将现实的原始数据处理并转化为数字图表，然后对其进行评估，并由此确定正确的行动路径。控制论协同计划散发着一种绝对可靠的气息，尽管像图灵这样的发明家知道计算机不可能如此完美地运行。正如控制论先驱斯塔福德·比尔（Stafford Beer）所言，我们倾向于使用机器来自动化现有的结构和流程，而这些结构和流程最初都是人类的创造。“我们将手、眼和大脑的局限性铭刻在钢铁、玻璃和半导体中，而计算机的发明正是为了超越这些局限性，”比尔在其1968年的著作《管理科学》（Management Sciences）中写道，并一针见血地指出了这一悖论。正如土耳其机器人（Mechanical Turk）一样，人类在机器中得以延续。

There remains an undeniable appeal to the photographs of Project Cybersyn. They appear over and over in design mood boards, projecting an aesthetic that still looks like the future many decades later. Perhaps the images are so influential because we retain that dream of the raw data of reality processed and crunched into digital graphs, which are then evaluated and, from there, the correct path of action determined. Project Cybersyn exuded an air of infallibility, even though inventors like Turing knew computers couldn’t work so perfectly. As the cybernetics pioneer Stafford Beer argued, we tend to use machines to automate the structures and processes that already exist, which were human creations to begin with. “We enshrine in steel, glass, and semiconductors those very limitations of hand, eye, and brain that the computer was invented precisely to transcend,” Beer wrote in his 1968 book Management Sciences, pinpointing the paradox. As with the Mechanical Turk, the human persists within the machine.

如今，我们确实拥有算法政府和算法生活的版本：银行使用机器学习来决定谁能获得贷款；Spotify 使用你过去的行为数据来推荐他们认为最符合你情感的歌曲。但实现这些壮举的技术并不像“控制协同计划”（Project Cyber​​syn）。那里没有六角形的房间，也没有高背椅。算法既无形又无处不在，包含在我们随身携带的手机应用程序中，即使它们的数据物理上托管在遥远的某个地方，在自然景观中不起眼的、巨大的空调服务器群中。“控制协同计划”认为，由数据运行的世界可能是连贯的、可理解的，包含在一个房间里，而我们现在知道它是抽象的、分散的，无处不在，又无处不在。我们被鼓励忘记算法的存在。

Today we do have versions of algorithmic government and algorithmic life: banks use machine learning to dictate who receives loans; Spotify uses the data of your past actions to determine songs to recommend, those they deem most aligned with your sensibility. But the technology that accomplishes those feats doesn’t look like Project Cybersyn. There are no hexagonal rooms or wingback chairs. Algorithms have become both invisible and omnipresent, contained in the apps we carry around with us on our phones even as their data are hosted physically somewhere distant, within vast air-conditioned server farms set into obscure locations in the natural landscape. Where Project Cybersyn suggested that the world run by data might be coherent and graspable, contained within a room, we now know that it is abstract and diffuse, everywhere and nowhere at once. We’re encouraged to forget the presence of algorithms.

新技术不可避免地会创造新的行为形式，但这些行为很少符合发明者预期。技术本身具有其内在的意义，最终会凸显出来。马歇尔·麦克卢汉在其1964年出版的《理解媒体：延伸》一书中，写下了著名的格言“媒介即信息”。人类的。他的意思是，新媒介——电灯、电话、电视——的结构比通过其传输的内容更重要。电话连接人们的能力超越了任何特定的对话。麦克卢汉写道：“任何媒介或技术的‘信息’都在于它为人类事务带来的规模、速度或模式的改变。” 在我们看来，媒介就是算法信息流；它以难以想象的程度扩展和加速了人类在世界各地的互联互通。它的信息是，在某种程度上，我们的集体消费习惯，转化为数据，汇聚成同质化。

New technologies inevitably create new forms of behavior, but the behaviors are rarely those that the inventors expect. The technology has an inherent meaning of its own that eventually comes to the fore. Marshall McLuhan wrote his famous dictum “the medium is the message” in his 1964 book Understanding Media: The Extensions of Man. He meant that the structure of a new medium—electric light, the telephone, television—is more important than the content that travels through it. The telephone’s ability to connect people exceeds any particular conversation. “The ‘message’ of any medium or technology is the change of scale or pace or pattern that it introduces into human affairs,” McLuhan wrote. In our case, the medium is the algorithmic feed; it has scaled and sped up humanity’s interconnection across the world to an unimaginable degree. Its message is that on some level, our collective consumption habits, translated into data, run together into sameness.

推荐算法的工作原理

HOW RECOMMENDATION ALGORITHMS WORK

算法是将一系列输入转化为特定输出的数字机器，就像工厂里的传送带一样。一种算法与其他算法的区别不在于其结构，而在于其构建的要素。所有推荐算法都通过收集一组原始数据来工作。该数据集的统称是信号，即收集到的输入数据，这些数据被输入到机器中。信号数据可能包括用户过去在亚马逊上的购买记录，或者有多少其他用户在 Spotify 上收藏了某首歌曲。数据是定量的而非定性的，因为它必须能够被机器处理。因此，即使数据是关于像音乐偏好这样主观的东西，它也会转化为数字：x个用户对y乐队的平均评分为z，或者x个用户听了y乐队的z次。许多社交媒体推荐中输入的主要信号是参与度，它描述了用户如何与内容互动。这可能以点赞、转发或播放的形式出现——任何一种在帖子旁边的按钮。高参与度意味着点赞、浏览或分享的数量高于其他帖子的平均值。

Algorithms are digital machines that turn a series of inputs into a particular output, like a conveyor belt in a factory. What makes one algorithm different from another is less their structure than the ingredients they are built from. All recommendation algorithms work by gathering a set of raw data. The overall term for that dataset is signal, the collected inputs that are fed into the machine. The signal data might include a user’s past purchases on Amazon or how many other users favorited a particular song on Spotify. The data is quantitative rather than qualitative, since it must be able to be processed by the machine. So even if the data is about something as subjective as music preferences, it is translated into numbers: x number of users rated y band an average of z, or x number of users listened to y band z times. The primary signal fed into many social media recommendations is engagement, which describes how users are interacting with a piece of content. That might come in the form of likes, retweets, or plays—any kind of button found next to a post. High engagement means the number of likes, views, or shares is higher than the average of other posts.

信号通过数据转换器传输，被转换成可用的数据包，并由不同类型的算法进行处理。参与度数据可能需要与评分数据或内容本身主题的数据分离。社交计算器可以用来添加关于用户在单一平台内如何相互关联的信息——我经常使用 Instagram例如，我的朋友安德鲁 (Andrew) 的帖子，这将使推荐系统更有可能在我的个人推送中将他的某个帖子排在靠前的位置。

The signal is fed through a data transformer that puts it into usable packages, set to be processed by different kinds of algorithms. Engagement data might need to be separated from ratings data, or data about the subject matter of the content itself. A social calculator might be used to add information about how users relate to one another within a single platform—I often engage with Instagram posts from my friend Andrew, for example, which would make a recommender system more likely to rank one of his posts highly in my personal feed.

然后是单个算法的具体方程。在当今的平台上，很少只有一套算法——而是有很多种。我们正在经历一系列不同的方程，这些方程考虑数据变量并以几种方式处理它们。一个方程仅根据参与度计算结果，也许找到平均参与度最高的内容，而另一个方程则优先考虑特定用户的内容的社交背景。这些算法也会相互权衡。混合过滤是指使用多种技术。最后，输出是推荐本身、自动播放列表中的下一首歌曲或有序的帖子列表。例如，算法会决定是否应该将朋友的生活更新放在 Facebook 动态中，而不是政治新闻报道上。

Then comes the specific equation of an individual algorithm. In today’s platforms, there is very rarely only one set algorithm—there are many. What we are experiencing is a series of different equations that consider data variables and process them in a few ways. One equation calculates a result based on engagement alone, perhaps finding the content with the highest average engagement, while another prioritizes the social context of a piece of content for a particular user. Those algorithms are also weighed against each other. Hybrid filtering is when multiple techniques are used. Finally, the output is the recommendation itself, the next song in the automated playlist or the ordered list of posts. The algorithm decides whether it should put a life update from a friend in your Facebook feed over a politics news story, for example.

音乐编目和推荐服务公司 Pandora 的一位高管曾向我描述，该公司的系统就像一个由算法组成的“管弦乐队”，其中还包含一个“指挥”算法。每个算法都使用不同的策略来提出推荐，然后指挥算法决定在特定时刻使用哪些推荐。（唯一的输出是播放列表中下一首要播放的歌曲。）不同的时刻需要不同的算法推荐技术。

An executive at the music cataloging and recommendation service Pandora once described the company’s system to me as an “orchestra” of algorithms, complete with a “conductor” algorithm. Each algorithm used different strategies to come up with a recommendation, and then the conductor algorithm dictated which suggestions were used at a given moment. (The only output was the next song to play in a playlist.) Different moments called for different algorithmic recommendation techniques.

不存在单一、不可或缺的“算法”，因为每个平台都有自己的运作方式，包含定制设计的变量和方程组。务必记住，Facebook 信息流的运作方式是一项商业决策，就像食品制造商决定使用哪种配料一样。算法也会随着时间推移而变化，利用机器学习不断完善自身。它们收集的数据用于逐步自我改进，以鼓励更多用户参与；机器适应用户，用户也适应机器。进入 2010 年代中期，随着社交媒体和流媒体服务加倍重视算法信息流，并开始主导用户体验，平台之间的差异变得更加突出和重要。

There is no single, monolithic “algorithm,” because each platform works in its own way, incorporating custom-designed variables and sets of equations. It’s important to remember that how the Facebook feed works is a commercial decision, the same as a food manufacturer deciding which ingredients to use. Algorithms also change over time, refining themselves using machine learning. The data they take in is used for gradual self-improvement to encourage even more engagement; the machine adapts to users and users adapt to the machine. The differences between platforms became more prominent and more relevant moving into the mid-2010s, as social media and streaming services doubled down on algorithmic feeds and they began to dominate the user experience.

我们用户从根本上并不了解算法推荐在日常运作中是如何运作的。它们的方程式、变量和权重并非公开，因为科技公司几乎没有动力去公开它们。它们是严格保密的商业机密，几乎就像核密码一样，对企业至关重要，很少被披露或提及。其中一个原因是，如果算法公开，用户就可以利用系统来推广自己的内容。另一个原因是对竞争的担忧：其他数字平台可能会窃取这些秘密，并制造出更好的产品。然而，这些工具与许多数字技术一样，最初都是在非商业环境下诞生的。

We users fundamentally do not understand how algorithmic recommendations work on a day-to-day basis. Their equations, variables, and weights are not public because technology companies have little incentive to publicize them. They are closely held trade secrets, almost like nuclear codes for how important they are to the businesses, and are rarely disclosed or hinted at. One reason for that is if the algorithms were public, users could game the system to promote their own content. Another is the fear of competition: other digital platforms could steal the secret sauce and make a better product. Yet these tools, like many digital technologies, started out in a non-commercial context.

推荐算法作为一种自动处理和排序信息的方式，在20世纪90年代开始付诸实践。最早的例子之一是电子邮件排序系统——至今仍是一项令人厌烦的苦差事。甚至在1992年，施乐公司帕洛阿尔托研究中心（更广为人知的名字是PARC）的工程师们就已经不堪重负。他们试图解决“电子邮件使用量日益增长，导致用户被大量涌入的文档淹没”的问题，David Goldberg、David Nichols、Brian M. Oki和Douglas Terry在1992年的一篇论文中写道。（他们当时并不知道我们在21世纪将面临多么巨大的数字通信量。）他们的电子邮件过滤系统Tapestry结合使用了两种算法：“基于内容的过滤”和“协同过滤”。前者已经应用于多个电子邮件系统，它会评估电子邮件的文本——比如，如果你想对所有包含“算法”的内容进行优先级排序。但后者更具创新性的技术是基于其他用户的行为。谁打开了某封邮件以及他们如何回复，都会被纳入系统对该邮件的优先级排序中。正如论文中所描述的：

Recommendation algorithms as a way of automatically processing and sorting information were put into practice in the 1990s. One of the first examples was a system for sorting email—to this day, an annoying chore. Even in 1992, engineers at Xerox’s Palo Alto Research Center (better known as PARC) were already overwhelmed by it. They sought to solve the problem of “the increasing use of electronic mail, which is resulting in users being inundated by a huge stream of Incoming documents,” David Goldberg, David Nichols, Brian M. Oki, and Douglas Terry wrote in a 1992 paper. (Little did they know the volume of digital communication we would face in the twenty-first century.) Their email filtering system, called Tapestry, used two kinds of algorithms in tandem: “content-based filtering” and “collaborative filtering.” The former, which was already used in several email systems, evaluated the text of emails—say, if you wanted to prioritize everything with the word algorithm. But the latter, more innovative technique was based on the actions of other users. Who opened a particular email and how they responded to it would be factored into how much the system prioritized the email. As the paper described it:

人们通过记录自己对所读文档的反应，相互协作，帮助彼此进行过滤。这些反应可能是：某个文档特别有趣（或特别无趣）。这些反应通常被称为注释，可以被其他人的过滤器访问。

People collaborate to help one another perform filtering by recording their reactions to documents they read. Such reactions may be that a document was particularly interesting (or particularly uninteresting). These reactions, more generally called annotations, can be accessed by others’ filters.

Tapestry 使用一个“过滤器”对一组文档进行重复查询，一个“小盒子”收集用户可能感兴趣的内容，以及一个“评估器”对文档进行优先级排序和分类。从概念上讲，它与我们现在看到的算法信息流非常相似：Tapestry 的目标是展示对用户最可能重要的内容。但该系统要求用户进行更多的前期操作，他们必须编写查询来确定自己想要查看的内容，查询基于内容或其他用户的操作。系统中的其他用户也必须进行非常有针对性的操作，依次将内容标记为引人注目或不相关。这样的系统需要一小群彼此熟悉并了解其群组如何与电子邮件互动的人——例如，你可能需要提前知道 Jeff 只会回复特别重要的电子邮件，因此你希望你的过滤器能够展示 Jeff 回复的所有电子邮件。Tapestry 在非常私密的环境中运行效果最佳。

Tapestry used a “filterer” to run repeated queries over a set of documents, a “little box” that collected material that might be of interest to the user, and an “appraiser” that could prioritize and categorize documents. Conceptually, it’s very similar to the algorithmic feeds we see now: Tapestry’s goal was to surface the content that was most likely to be important to the user. But this system required much more up-front action on the part of the user, who had to write queries to determine what they wanted to see, based either on content or on other users’ actions. The other users in the system also had to carry out very intentional actions, marking material as compelling or irrelevant in turn. Such a system required a small group of people who already knew one another and understood how their cohort interacted with email—for instance, you may need to be aware in advance that Jeff replies only to particularly important emails, so you want your filter to surface all the emails that Jeff replies to. Tapestry functioned best on a very intimate scale.

1995 年，麻省理工学院媒体实验室的 Upendra Shardanand 和 Pattie Maes 在一篇论文中描述了“社会信息过滤”，即“一种根据用户与其他用户的兴趣概况之间的相似性，从任何类型的数据库中向用户提供个性化推荐的技术”。该技术以 Tapestry 的理念为基础，是对网上信息过载的回应：“信息量远远超出了任何人为了找到自己喜欢的信息而能够筛选的范围。”他们得出结论，自动过滤器是必要的：“我们需要技术来帮助我们筛选所有信息，找到我们真正想要和需要的东西，并摆脱那些我们不想被打扰的东西。”（当然，这在网上仍然是一个大问题。）Shardanand 和 Maes 认为基于内容的过滤存在重大缺陷。它需要将材料转换成机器可以理解的数据，例如文本；它缺乏偶然性，因为它只能根据用户输入的术语进行过滤；而且它无法衡量内在质量。它无法“区分一篇写得好的文章和一篇写得不好的文章，如果两篇文章使用了相同的术语”。这种无法评估质量的能力让人想起了人工智能：像ChatGPT这样的新工具似乎能够理解并生成有意义的语言，但实际上，它们只是重复训练所用现有数据中固有的模式。质量是主观的；在缺乏人类判断的情况下，单凭数据只能在一定程度上衡量质量。

In 1995, a paper from Upendra Shardanand and Pattie Maes at the MIT Media Lab described “social information filtering,” “a technique for making personalized recommendations from any type of database to a user based on similarities between the interest profile of that user and those of other users.” Building on the ideas of Tapestry, it was a response to the overflow of information online: “The volume of things is considerably more than any person can possibly filter through in order to find the ones that he or she will like.” Automated filters would be necessary, they concluded: “We need technology to help us wade through all the information to find the items we really want and need, and to rid us of the things we do not want to be bothered with.” (Of course, this is still a huge problem online.) Shardanand and Maes argued that content-based filtering had significant drawbacks. It requires the material to be translated into data that the machine can understand, such as text; it lacks serendipity because it can filter only by the terms that the user inputs; and it does not measure inherent quality. It is unable to “distinguish a well written [and] a badly written article if the two articles use the same terms.” The inability to evaluate quality brings to mind artificial intelligence: New tools like ChatGPT seem to be able to understand and generate meaningful language, but really, they only repeat patterns inherent in the preexisting data they are trained on. Quality is subjective; data alone, in the absence of human judgment, can go only so far in gauging it.

社交信息过滤绕过了这些问题，因为它是由人类用户的行为驱动的，人类用户会自行评估内容——运用定量和定性的判断。它更像是口口相传，就像我们从与我们偏好相似的朋友那里获得听什么或看什么的建议一样：“根据其他品味相似的人给出的评分，系统会向用户推荐内容。” 论文指出。用户品味的相似度是通过统计相关性计算得出的。研究人员设计了一个名为 Ringo 的系统，使用电子邮件列表进行音乐推荐。当用户评估音乐时，系统会根据 1 到 7 的等级对初始批次的 125 位艺术家进行评分，并构建他们的偏好图表。然后，通过将该图表与其他用户的偏好图表进行比较，系统会推荐他们可能喜欢的音乐——或者讨厌的音乐（这也是一种选择）。Ringo 的推荐带有置信度指标，表明建议的正确性，并允许用户进一步考虑算法的选择。到 1994 年 9 月，Ringo 已拥有 2100 名用户，每天收到 500 封评价音乐的电子邮件。

Social information filtering bypasses those problems because it is instead driven by the actions of human users, who evaluate content on their own—using judgments both quantitative and qualitative. It’s more like word of mouth, the way we get advice on what to listen to or watch from friends whose preferences are similar to our own: “Items are recommended to a user based upon values assigned by other people with similar taste,” according to the paper. The similarity of one user’s taste to another was calculated using statistical correlation. The researchers designed a system called Ringo to make music recommendations using an email list. As a user evaluated music, rating an initial batch of 125 artists on a scale of 1 to 7, a diagram of their preferences was built. Then, by comparing that diagram to other users’, the system suggested music that they were likely to enjoy—or hate, which was also an option. Ringo recommendations came with a measure of confidence, signaling how likely a suggestion was to be right and allowing the user to further consider the algorithmic choice. By September 1994, Ringo had twenty-one hundred users and five hundred emails a day evaluating music.

Ringo 测试了各种特定的算法，以便根据音乐评分做出决策。第一种算法测量用户品味的差异性，并根据最相似的用户推荐音乐。第二种算法测量相似性，然后利用与其他用户的正负相关性做出决策。第三种算法确定不同艺术家之间的相关性，推荐与用户已经喜欢的艺术家高度相关的艺术家。第四种算法，也是研究人员认为最有效的算法，根据用户对同一事物的正面或负面评价来匹配用户。换句话说，他们的品味匹配。相似性是最好的变量。系统中的用户越多，用户预先输入的信息越多，Ringo 的效果就越好——一些用户甚至将其描述为“令人不安的准确”。Ringo 的创新之处在于它承认最好的推荐，或者说最佳的相关性指示，很可能来自其他人，而不是对内容本身的分析。这代表着人类品味的提升。

Ringo tested various specific algorithms to make decisions based on the music ratings. The first algorithm measured dissimilarity between users’ tastes and based recommendations on the most similar users. The second algorithm measured similarity, then used positive and negative correlations with other users to make decisions. The third algorithm determined the correlation between different artists, recommending artists who were strongly correlated to those a user already liked. The fourth algorithm, and the most effective according to the researchers, matched users based on whether they rated the same things either positive or negatively. In other words, their taste matched. Similarity was the best variable. The more users in the system, and the more input users gave up front, the better Ringo worked—some users even described it as “unnervingly accurate.” Ringo’s innovation was how it acknowledged that the best recommendations, or the best indications of relevance, were likely to come from other humans rather than analysis of the content itself. It represented a scaling up of human taste.

早期的互联网算法旨在从海量信息中筛选出对用户重要的内容，并以连贯的方式呈现。推荐是其目标：推荐一条信息、一首歌曲、一张图片或一条社交媒体动态。算法推送有时被更正式、更直观地称为“推荐系统”，指的是选择内容的简单操作。

Early Internet algorithms were designed to sift through a vast body of material for whatever was important to a user, and then present it in a coherent way. Recommendations were the goal: recommending a piece of information, a song, an image, or a social media update. Algorithmic feeds are sometimes more formally and literally labeled “recommender systems,” for the simple act of choosing a piece of content.

第一个完全主流的互联网算法，几乎每个互联网用户都遇到过，是谷歌搜索算法。1996年，谷歌联合创始人谢尔盖·布林和拉里·佩奇在斯坦福大学学习期间，开始研究后来成为PageRank的系统。该系统用于抓取互联网（当时互联网总量约为一亿份文档），并识别哪些网站和页面比其他网站更有用或更具信息量。PageRank的工作原理是测量一个网站被其他网站链接的次数，类似于学术论文引用过去研究关键内容的方式。链接越多，页面可能就越重要。布林和佩奇在1998年的论文《大型超文本网络搜索引擎的剖析》中写道，引用量指标“与人们对重要性的主观看法非常吻合”。PageRank将协同过滤与内容过滤相结合。通过链接不同的页面，人类用户已经形成了一个主观的推荐图谱，该算法可以将其纳入其中。它还衡量了页面上的链接数量、链接的相对质量，甚至文本的大小等因素——文本越大，与特定搜索词的相关性就越高。PageRank 值越高的页面，就越有可能出现在 Google 搜索结果列表的顶部。

The first wholly mainstream Internet algorithm, one that almost every Internet user has encountered, was the Google Search algorithm. In 1996, while studying at Stanford University, Sergey Brin and Larry Page, the cofounders of Google, began work on what would become PageRank, a system for crawling the Internet (which at that point amounted to perhaps one hundred million documents in total) and identifying which sites and pages were more useful or informative than others. PageRank worked by measuring how many times a website was linked to by other sites, similar to the way academic papers cite key pieces of past research. The more links, the more important a page was likely to be. The metric of citation “corresponds well with people’s subjective idea of importance,” Brin and Page wrote in a 1998 paper, “The Anatomy of a Large-Scale Hypertextual Web Search Engine.” PageRank mingled a form of collaborative filtering with content filtering. By linking various pages, human users had already formed a subjective map of recommendations that the algorithm could incorporate. It also measured factors like the number of links on a page, the relative quality of the links, and even the size of text—the larger the text, the more relevant it might be for a particular search term. Pages with a higher PageRank were more likely to appear at the top of Google’s list of search results.

佩奇和布林预测他们的系统将随着互联网的发展而保持功能性和可扩展性，这一预测是正确的。几十年后，PageRank 几乎成了一个专制的系统，它主宰着网站如何以及何时被浏览。对于企业或资源而言，适应并适应 Google 搜索结果首页至关重要。PageRank 算法。21 世纪初，为了找到我想要的内容，我得连续浏览好几页谷歌搜索结果。最近，我几乎再也找不到第二页了，部分原因是谷歌搜索现在会预先加载它认为相关的文本，从网站上提取文本，并在搜索页面顶部直接显示给用户，而不是实际结果。因此，像“我可以给我的狗喂胡萝卜吗？”这样的问题——在我养狗初期，我经常在谷歌上搜索——用户无需加载其他网站就能得到答案，这进一步巩固了谷歌的权威。“知识本身就是力量”，弗朗西斯·培根在 16 世纪写道，但在互联网时代，对知识进行分类可能更有力量。现在，信息很容易找到，但理解信息，知道哪些信息有用，却要困难得多。

Page and Brin’s prediction that their system would remain functional and scalable as the Internet grew were correct. Decades later, PageRank has become almost tyrannical, a system that dominates how and when websites are seen. It’s vital for a business or resource to make it to that first page of Google Search results by adapting to the PageRank algorithm. In the early 2000s, I perused many successive pages of Google results to find exactly what I was looking for. More recently, I hardly ever make it to the second page, in part because Google Search now frontloads text that it gauges will be relevant, pulling it from websites and displaying it directly to the user at the top of the search page, before the actual results. Thus a query like “Can I feed my dog carrots?”—the kind of question I googled incessantly in the early days of puppy ownership—will deliver an answer without a user ever having to load another site, further consolidating Google’s authority. “Knowledge itself is power,” Francis Bacon wrote in the sixteenth century, but in the Internet era, sorting knowledge might be even more powerful. Information is now easy to find in abundance; making sense of it, knowing which information is useful, is much harder.

佩奇和布林希望他们的系统相对中立，仅根据相关性来评估每个网站。算法的指令是优先为用户提供最佳信息。迎合特定网站或企业会破坏搜索结果。“我们预计，由广告资助的搜索引擎将本质上偏向广告商，而远离消费者的需求，”两位企业家在1998年写道。然而，在2000年，他们推出了谷歌AdWords，作为公司面向广告商的试点产品。如今阅读他们的批评颇具趣味，因为广告如今已成为谷歌绝大部分收入来源——2020年超过80%。由于PageRank算法吸引了数十亿用户使用谷歌搜索，该公司还可以追踪用户的搜索内容，从而可以向广告商出售特定搜索查询的广告位。用户看到的广告与搜索结果一样，都受到算法的影响。而建立在搜索算法之上的广告，将谷歌变成了一个庞然大物。

Page and Brin wanted their system to be relatively neutral, evaluating each site solely in terms of its relevance. The algorithm’s directive was to prioritize the best information for the user. Catering the search to a particular site or business would ruin the results. “We expect that advertising funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers,” the entrepreneurs wrote in 1998. Yet, in 2000, they launched Google AdWords as the company’s pilot product for advertisers. It is amusing to read their critique today, as advertising now provides the vast majority of Google’s revenue—more than 80 percent in 2020. As PageRank attracted billions of users to Google Search, the company could also track what the users were searching for and could thus sell advertisers space on particular search queries. The ads a user sees were just as informed by the algorithm as the search results were. And advertising, built on the search algorithm, turned Google into a behemoth.

到21世纪初，算法过滤已经主宰了我们的数字体验。亚马逊网站早在1998年就开始使用协同过滤技术为顾客推荐产品。该系统不像Ringo那样，试图通过测量相似用户的个人资料来大致了解他们的喜好，而是通过确定哪些商品可能会被同时购买——这让人很恼火。例如婴儿奶瓶。2017 年，亚马逊一名员工与他人合著的一篇论文描述了网站上充斥着此类建议的情况：

By the early 2000s, algorithmic filtering was already dictating our digital experiences. The Amazon website began using collaborative filtering as early as 1998 to recommend products for customers to buy. Rather than attempting to measure similar profiles of users to approximate taste, as Ringo did, the system worked by determining which items were likely to be purchased in tandem—a rattle with a baby bottle, for example. A 2017 paper cowritten by an Amazon employee described the bombardment of such suggestions on the site:

主页醒目地展示了根据您过去的购买记录和在商店浏览过的商品提供的推荐……购物车会推荐其他可添加到购物车的商品，可能是最后一刻冲动购买的商品，也可能是与您已经在考虑的商品互补的商品。在您下单后，还会显示更多推荐，建议稍后再订购。

The homepage prominently featured recommendations based on your past purchases and items browsed in the store…. The shopping cart recommended other items to add to your cart, perhaps impulse buys to bundle in at the last minute, or perhaps complements to what you were already considering. At the end of your order, more recommendations appeared, suggesting items to order later.

算法推荐就像Trader Joe's超市收银台前的货架，最后推送你可能需要的商品。但在这种情况下，推荐的内容是针对每个网站用户量身定制的，正如论文所述，最终实现了“每位顾客都能拥有一家商店”。亚马逊发现，个性化的产品推荐在点击率和销售额方面比横幅广告和畅销产品列表等非个性化营销策略更有效，因为后者的针对性不够强。推荐算法提升了业务，也为顾客带来了便利，他们可能会找到一些他们原本不知道自己需要的东西。（目前，我的亚马逊主页推荐了一款无线高压清洗机和一款日式煎蛋锅。）

The algorithmic recommendations resemble the shelves stationed just before the register in a Trader Joe’s, one last push of products that you may need. But in this case, what was recommended was tailored to each website user, resulting in “a store for every customer,” as the paper described. Amazon found that the personalized product recommendations were much more effective in terms of click-throughs and sales than unpersonalized marketing tactics like banner advertisements and lists of bestselling products, which can’t be as tightly targeted. The recommendation algorithm improved business and appeared convenient for the customer, who might find something they didn’t know they needed. (Right now, my Amazon home page recommends a cordless power washer and a Japanese omelet pan.)

这些早期算法对个人电子邮件、音乐家（而非特定歌曲）、网页和商业产品进行分类。随着数字平台的扩张，推荐系统进入了更复杂的文化领域，运行速度更快，数据量更大，能够对数百万条推文、电影、用户上传的视频，甚至潜在的恋爱对象进行分类。筛选成为了默认的在线体验。

These early algorithms sorted individual emails, musicians (as opposed to specific songs), web pages, and commercial products. As digital platforms expanded, recommender systems moved into more complex areas of culture and operated at much faster speeds and higher volumes, sorting millions of tweets, films, user-uploaded videos, and even potential romantic partners. Filtering became the default online experience.

这段历史也提醒我们，推荐系统并非无所不知，而是由一群技术研究人员或工作者构建的工具。它们是会犯错的产品。尼克·西弗是一位社会学家，也是塔夫茨大学研究推荐系统的教授。他的研究重点是算法的人性化方面，即如何工程师们让他们思考算法推荐。在我与他的讨论中，西弗总是确保澄清算法的模糊性，将个体方程与其设计背后的企业动机及其最终对用户的影响区分开来。“算法是公司整体的转喻，”他告诉我。“Facebook 算法并不存在；Facebook 存在。算法是谈论 Facebook 决策的一种方式。”

This history is also a reminder that recommender systems are not omniscient entities but tools built by groups of tech researchers or workers. They are fallible products. Nick Seaver is a sociologist and a professor at Tufts University who studies recommender systems. His research focuses on the human side of algorithms, how the engineers who make them think about algorithmic recommendations. In my discussions with him, Seaver always made sure to clarify the ambiguous entity of the algorithm, separating the individual equation from the corporate motives behind its design and its eventual impact on the user. “The algorithm is metonymic for companies as a whole,” he told me. “The Facebook algorithm doesn’t exist; Facebook exists. The algorithm is a way of talking about Facebook’s decisions.”

技术本身并非问题所在——我们既不能将糟糕的推荐归咎于算法本身，也不能将工程缺陷归咎于桥梁。为了使数字平台上海量的内容更容易被理解，某种程度的重新排序是必要的。Filterworld 的负面影响或许源于该技术被过度广泛应用，而没有充分考虑用户的体验，而不是针对目标用户的广告商。这些推荐系统，即便如此，对我们来说也不再有效；相反，我们对它们的疏远程度正在日益加深。

The technology is not at issue—one can no more blame an algorithm itself for bad recommendations than blame a bridge for its engineering flaws. And some degree of reordering is necessary to make the vast stores of content on digital platforms comprehensible. The negative aspects of Filterworld might have emerged because the technology has been applied too widely, without enough consideration for the experience of the user, rather than for the advertisers targeting them. The recommendations, such as they are, don’t work for us anymore; rather, we are increasingly alienated by them.

早期社交媒体

EARLY SOCIAL MEDIA

我对社交媒体最早有意义的记忆来自 Facebook。我在接受塔夫茨大学的录取通知后加入了这个平台。当时是 2006 年夏天，潜在用户需要一个官方的 .edu 电子邮件地址才能访问该平台的完整大学部分。与现在相比，Facebook 的早期版本几乎无法辨认。那时，Facebook 的覆盖范围非常有限；我主要用它与其他即将入学的塔夫茨学生联系。如果说今天的 Facebook 是一条每隔几秒钟就有出口和入口的疯狂高速公路，那么在 21 世纪初，它更像是一个高中娱乐室，一次只能容纳几个人闲逛。你可以创建个人资料，在个人资料上更新你的状态，并加入具有共同兴趣的小组——但仅此而已。

My first meaningful memories of social media come from Facebook, which I joined after I accepted an admission offer from Tufts University, where I went to college. At that time, in summer 2006, prospective users needed an official .edu email address to access the full college section of the platform. That early iteration of Facebook is nearly unrecognizable when compared to its present-day anatomy. Back then, Facebook’s reach was strictly limited; I mainly used it as a means to connect with other incoming Tufts students. If Facebook today is a frenetic highway with exits and on-ramps every few seconds, in the aughts it was more like a high school rec room where only a few people could hang out at a time. You built a profile, updated your status on the profile, and joined groups around common interests—but not much else.

Facebook 并非最早的在线社交方式。Friendster 和 MySpace 才是它的前身。AOL 的即时通讯工具和谷歌的 gChat 提供了引人入胜的实时与好友交流的方式。到 2006 年，我已经在 Facebook 上花了数百个小时在讨论电子游戏和音乐的旧论坛网站上，扎克伯格的Facebook将线上身份与线下世界紧密联系在一起。该平台鼓励用户使用真实姓名，而不是晦涩难懂的化名，并影响了大学这个小圈子里的现实生活计划：举办派对、规划学术活动以及发展人际关系。由此，它为数百万乃至数十亿用户的在线社交生活主流化铺平了道路。

Facebook was hardly the first way to socialize online. Friendster and MySpace were its predecessors. AOL’s Instant Messenger and Google’s gChat provided engrossing ways to hang out with your friends in real time. By 2006 I had already spent hundreds of hours on older forum websites discussing video games and music. But Zuckerberg’s Facebook tied online identity coherently and consistently to the offline world. The platform encouraged users to use their real names rather than arcane aliases and influenced real-life plans in the small world of college: throwing parties, planning academic activities, and conducting relationships. In doing so, it paved the way for the mainstreaming of online social life for millions, and then billions, of users.

2006年9月，在我加入Facebook后不久，Facebook实施了其最大的变革之一，这项功能奠定了它未来成为互联网上“货比三家”的格局。新闻推送（News Feed）——一个不断更新、帖子和提醒的列表——成为了该平台的核心功能。它就像一条穿过宁静村庄的新建高速公路，令人无法忽视。Facebook官方更新公告写道：“现在，无论何时登录，你都能获得好友和社交群组活动生成的最新头条新闻。”

In September 2006, not long after I joined the network, Facebook implemented one of its biggest changes, a feature that would set the course for its future as the big-box everything-for-sale store of the Internet. The News Feed, a running list of updates, posts, and alerts, became the primary feature of the platform. It was unignorable, like a newly built highway that cut through a quiet village. “Now, whenever you log in, you’ll get the latest headlines generated by the activity of your friends and social groups,” Facebook’s official update note announced.

同年提交的新闻推送专利（尽管直到2012年才获得批准）描述了其用途：“一种系统和方法，为在社交网络环境中使用电子设备的用户提供动态选择的媒体内容。” 换句话说，新闻推送是由算法决定向用户显示什么内容的信息流。另一项专利声称能够“生成基于关系的动态内容，为基于网络的社交网络的成员提供个性化服务”。起初，新闻推送只是一连串关于约会状态变化和个人资料图片更新的公告。它并没有特别的威胁性。

The patent for the News Feed, filed that year, though it wasn’t granted until 2012, described its purpose: “A system and method provides dynamically selected media content to someone using an electronic device in a social network environment.” In other words, the News Feed was a flow of information dictated by an algorithm that determined what to show a user. Another patent claimed the ability to “generate dynamic relationship-based content personalized for the members of the web-based social network.” At first the News Feed was just a stream of announcements of changed dating statuses and updated profile pictures. It wasn’t particularly threatening.

新闻推送专利申请的详细描述表明，这是一种协同过滤系统，其规模远超 20 世纪 90 年代的电子邮件系统。它值得全文引用，因为它预测了从社交网络到流媒体和电子商务等网络生活的大部分内容在随后的十年里会变成什么样：大量自动化推送更多地由企业而非用户主导，逐渐形成了用户与内容推送之间更为被动的关系。

The News Feed patent application’s longer description suggests a system of collaborative filtering, acting on a much larger scale than the email systems of the 1990s. It’s worth quoting in full because it predicts what much of life online, from social networks to streaming and e-commerce, became in the decade that followed: so many automated feeds dictated by corporations more so than users, gradually forming a more passive relationship between users and the content feed.

根据用户与一个或多个其他用户的关系，为用户选择媒体内容。用户与其他用户的关系反映在所选媒体内容及其格式中。例如，根据媒体内容对用户的预期重要性，为媒体内容分配一个顺序，并按照分配的顺序向用户显示媒体内容。用户可以更改媒体内容的顺序。系统会监控用户与社交网络环境中可用媒体内容的交互，并根据这些交互为用户选择其他媒体内容。

Items of media content are selected for the user based on his or her relationships with one or more other users. The user’s relationships with other users are reflected in the selected media content and its format. An order is assigned to the items of media content, for example, based on their anticipated importance to the user, and the items of media content are displayed to the user in the assigned order. The user may change the order of the items of media content. The user’s interactions with media content available in the social network environment are monitored, and those interactions are used to select additional items of media content for the user.

这段文字体现了算法推送的所有要素——该系统通过监控用户过去参与的内容，预测某条内容对个人用户的相对重要性，并将被认为最有可能同样引人入胜的内容推送到列表顶部。其目标是筛选内容，选出最有趣的内容，从而鼓励用户消费更多内容，并关注更多账户。用户能够更频繁地使用社交媒体，并在网站上停留更长时间，这正是算法推送的可行性所在。（如果我们的朋友在Facebook上不活跃，就像我近年来的情况一样，那么我们的活动也可能会减少。）

All the elements of the algorithmic feed are present in this passage—a system that anticipates a piece of content’s relative importance to an individual user, determined by surveillance of content they engaged with in the past, and pushes whichever content is deemed most likely to be equally engaging to the top of the list. The goal was to filter content to select what is most interesting, therefore encouraging a user to consume a higher volume of content and follow more accounts overall. Users being able to use social media more often and stay on the sites longer is what made them viable. (If our friends aren’t active on Facebook, which has been the case for me in recent years, then we are likely to tail off our activity, too.)

起初，News Feed 纯粹按时间顺序排列，最新更新优先，但逐渐转向算法逻辑。随着 Facebook 的发展壮大，用户连接不断增加，从个人关系扩展到出版物和品牌，个人更新量也随之增加。随着时间的推移，更新内容不再只是来自朋友的日常留言，还包括群组消息、新闻链接和促销公告。普通用户无法指望按照时间顺序浏览如此海量且种类繁多的帖子，即使尝试，他们也会不知所措，或者错过重要的帖子，这可能会导致他们对平台感到不满。最终，消费的规模和速度使得 Facebook 有必要进行积极的算法过滤。

At first, the News Feed was ordered purely chronologically, with the most recent updates first, but it gradually followed a more algorithmic logic. As Facebook grew and users added more connections, expanding from personal relationships to publications and brands, the volume of individual updates increased. Over time, the updates weren’t just mundane notes from friends but messages from groups, links to news stories, and announcements of sales. Casual users couldn’t hope to follow a chronological feed with such a volume and variety of posts, and if they tried, they would either be overwhelmed or fail to catch an important post, which might cause dissatisfaction with the platform. Ultimately, the scale and speed of consumption made aggressive algorithmic filtering necessary for Facebook.

Facebook 的“赞”按钮及其标志性的竖起大拇指按钮于 2009 年推出，它提供了一种数据形式，表明用户对特定内容的兴趣程度。用户参与度以点赞、评论和一个帐户之前与另一个帐户的互动来衡量，并被纳入信息流的排序中。该算法系统称为 EdgeRank，Facebook 将其主要变量确定为亲和力分数、边缘权重和时间衰减。“边缘”指的是人们在 Facebook 上进行的任何操作，然后将其作为更新发送到新闻提要中列出。亲和力分数表示用户与发帖人的联系程度和联系强度（例如，持续评论朋友的帖子）。评论比点赞更重要，最近的互动比旧的互动更重要。边缘权重评估了不同类别的互动：朋友发布新照片的更新可能比发布新闻文章链接或加入新群组在算法中获得更大的权重。时间衰减是操作的年龄；如果其他因素相同，近期操作比旧操作更有可能出现在新闻推送顶部。EdgeRank 的得分并非像锦标赛中篮球比赛的结果那样一次性永久分配，而是会随时变化。这三个类别并非简单的单一、中立的数据点；它们是 Facebook 以特定方式打包和解读的数据集合。

Facebook’s Like button, with its signature thumbs-up, was introduced in 2009, providing one form of data on how interested a user might be in a particular piece of content. User engagement, measured by likes, comments, and one account’s previous interactions with another, factored into the order of the feed. That algorithmic system was called EdgeRank, and Facebook identified its principal variables as affinity score, edge weight, and time decay. “Edge” referred to any action people carry out on Facebook, which is then sent to the News Feed as an update to be listed. Affinity score represented how connected a user was to the poster and the strength of the connection (e.g., consistently commenting on friend’s posts). A comment counted more than a like, and recent interactions counted more than older ones. Edge weight evaluated different categories of interactions: an update of a friend posting a new photo might be given more weight by the algorithm than posting a link to a news article or joining a new group. Time decay was the age of the action; recent actions were more likely to be at the top of the News Feed than older ones, if the other factors were equal. The EdgeRank scores were not permanently assigned once, like the outcome of a basketball game in a tournament, but changed instant to instant. And those three categories aren’t simply single, neutral data points; they are collections of data packaged and interpreted in specific ways by Facebook.

由于 Facebook 算法信息流不断更新，且该公司只是断断续续地披露细节，因此很难追踪其演变。除了官方公告之外，我们对其的了解主要来自记者的调查报道和用户的体验，而这些用户早在算法更新公开之前就能感受到其影响。熟悉的网站在信息流机制发生变化时，总会给人一种不同的感觉。例如，在 Facebook 上，你可能会注意到好友的帖子越来越少，而来自群组或企业的帖子越来越多；或者 Instagram 上永远不会显示某个好友的帖子，因此你需要使用搜索栏才能找到他们。

It’s hard to track the evolution of Facebook’s algorithmic feed because it is constantly updated, and the company reveals details only intermittently. What we do know about it beyond official announcements comes down to investigative reporting from journalists and the experiences of users, who see the effects of updated algorithms long before they’re made public. Familiar websites have a way of feeling different when the feed mechanism changes. On Facebook, for instance, you may notice that you see less of your friends’ posts and more from groups or businesses, or that Instagram never shows you posts from a particular friend in your feed and you thus need to hunt them down using the search bar.

算法推送本身并非始终如一，也并非直线路径，最终趋向完美。它会随着公司优先级的变化而变化。2011年，Facebook 将新闻推送描述为“你自己的个人报纸”，表明其目标是将社交更新与外界新闻报道融合在一起。2013 年，该公司表示其算法可以“检测被定义为高质量的内容”。但在 2010 年代，追求公司衡量的“高质量”是一种荒谬的游戏。如果你想让你的 Facebook 帖子受到关注——这对新闻出版物和自由撰稿人来说是一个大问题——你必须猜测什么样的材料会被优先考虑。这种关系几乎是对立的；只有你“玩弄”算法，你才会被听到。你不能再依赖关注你或加你为好友的用户看到你的帖子了。

The algorithmic feed itself is not consistent or on a linear path toward some ultimate perfection. It changes with a company’s priorities. In 2011, Facebook described the News Feed as “your own personal newspaper,” suggesting its goal of mingling social updates with news stories from the outside world. In 2013, it said its algorithm worked to “detect content defined as high quality.” But chasing whatever the company gauged as “high quality” was something of an absurd game over the course of the 2010s. If you wanted to get attention for your Facebook posts—a big problem for journalism publications and freelance writers—you had to guess at what kind of material was getting prioritized. The relationship was almost oppositional; only if you “gamed” the algorithm would you be heard. You could no longer rely on users who had followed or friended you seeing your posts.

在我的自由新闻职业生涯中，我记得曾有传言说，文章链接不再受算法高度重视。因此，我和许多其他记者不再直接用简单的链接发布报道，而是只在帖子评论中添加链接。这种伎俩本意是为了促进算法的推广，尽管这更容易让读者感到困惑。后来，我发现，撰写类似结婚公告的文字和“祝贺”的评论，能把帖子推到信息流的顶部。于是我开始用虚假的婚礼或其他人生里程碑来分享我的文章。这些现象表明，当用户试图玩弄算法或逃避检测时，算法是如何扭曲语言本身的。最近，在TikTok上，出现了一些委婉语来表达那些触发算法屏蔽或减慢视频速度的词语：用“unalive”（不活着）来表示杀戮，“SA”（性侵犯），用“spicy eggplant”（辣茄子）来代替振动器，正如记者泰勒·洛伦兹在《华盛顿邮报》上记录的那样。此类词汇被昵称为“算法语言”：按照算法的形象塑造的言语。

At one point in my freelance journalism career, I recall a rumor going around that links to articles were no longer very highly weighted by the algorithm. So rather than posting our stories directly using a simple link, I and many other journalists added a link to the story only by commenting on the post. The trick was supposed to goose algorithmic promotion, even though it was more confusing for a reader. At another point, it became clear that writing text that resembled a marriage announcement and comments that said “congratulations” pushed posts to the top of the feed. So I began sharing my articles with fake weddings or other life milestones. These phenomena show how algorithms can warp language itself as users attempt to either game them or evade detection. More recently, on TikTok, euphemisms have emerged for terms that trigger the algorithm to block or slow down a video: “unalive” for kill, “SA” for sexual assault, “spicy eggplant” instead of vibrator, as the journalist Taylor Lorenz documented in the Washington Post. Such vocabulary was nicknamed “algospeak”: speech molded in the image of the algorithm.

我不清楚我在Facebook上使用的技巧是否真的有效，但为了接触潜在读者，我愿意尝试任何方法。这就像为谷歌搜索引擎优化设计网站：记者根据算法指标（至少是我们认为的指标）来优化内容。这个过程感觉像是在操控，有时甚至像卡夫卡式的；我们是在与一个看不见、摸不着、不断变化的对手较量。

It was unclear if the tricks I used on Facebook had much of an impact, but I was willing to try anything to reach potential readers. It was like designing a website for Google search-engine optimization: journalists optimized content for the metrics of the algorithm, or at least what we perceived them to be. The process felt manipulative and at times Kafkaesque; we contended with an unseen, incomprehensible, ever-changing opponent.

2015 年左右，Facebook 决定优先考虑视频内容，因此推荐算法对视频的推广力度远超以往。媒体公司随后“转向”制作视频来吸引这些受众，有时还会得到 Facebook 自身的资助。这种努力只持续了几年，Facebook 再次降低视频的优先级，导致包括 BuzzFeed、Mashable 和 MTV 在内的这些媒体公司掀起裁员潮。（据一项诉讼称，该项目结束后，Facebook 还被发现谎报了视频的流量，将数字夸大了九倍。）算法推送的内容一直在变化。2016 年，Facebook 为帖子添加了“回应”功能，这样观众可以使用各种表情符号来回应，而不仅仅是点赞按钮。收到很多表情符号回应的帖子得到了更多的推广。但这种改变也适得其反，因为煽动性内容（例如，收到很多愤怒表情回应的帖子，比如引发愤怒的政治故事）得到了过多的推广，并破坏了整个网站的基调。吸引了更多人的关注并不一定意味着这些帖子更有价值。

Around 2015, Facebook decided to prioritize video content, so the recommendation algorithm promoted videos much more than it did previously. Media companies then “pivoted” to making videos to chase that audience, sometimes with the help of funding from Facebook itself. That effort lasted only a few years, and then Facebook deprioritized videos once more, leading to waves of layoffs at those same media companies, including BuzzFeed, Mashable, and MTV. (After the program ended, it also emerged that Facebook had lied about the traffic the videos were getting, inflating the numbers up to nine times, according to a lawsuit.) The algorithmic feed kept shifting. In 2016, Facebook added “reactions” to posts, so that viewers could respond with a range of emoticons rather than just the Like button. Posts that received many emoticon reactions got more promotion. But that change backfired, too, when incendiary content—posts that received many angry-face reactions, for example, like rage-inducing political stories—was getting too much promotion and souring the tone of the entire site. That they attracted more engagement didn’t mean the posts were necessarily more worthwhile.

从按时间顺序排列的信息流转向日益增加的算法推荐的并非只有 Facebook。几乎所有主流社交网络在 2010 年代都走上了同样的道路。Filterworld 正是在 2010 年代中期算法化加剧的时期开始成形。

It wasn’t only Facebook that moved from a chronological feed to an increasing volume of algorithmic recommendations. Almost every major social network followed the same path over the 2010s. Filterworld began taking shape in the middle of the decade when algorithmification intensified.

Facebook 于 2012 年收购了 Instagram，当时它只有 13 名员工。此后的几年里，这款照片分享应用变得越来越像 Facebook 本身，从好友上传照片的线性信息流转变为视频、广告和推荐帖子流。2016 年 3 月，Instagram 的信息流开始从时间顺序转换为算法排列。这一变化首先在小部分用户群中进行测试，然后推广到越来越多的用户，直到影响到所有人。越来越无序的信息流引发了一种困惑和焦虑感，类似于有人在你不知情的情况下重新布置了你家里的家具的感觉。以前，通过滚动浏览信息流，你可以回到过去。但突然之间，两天前的帖子出现在了你的信息流顶部。

Facebook acquired Instagram in 2012, when it only had thirteen employees. In the years since, the photo-sharing app has become more like Facebook itself, moving away from a linear feed of photos uploaded by friends into a stream of videos, ads, and recommended posts. In March 2016, the Instagram feed began switching from a chronological to an algorithmic arrangement. The change was tested on small groups of users then rolled out to more and more, until it hit everyone. The increasingly out-of-order feed induced a sense of confusion and anxiety akin to the feeling of someone rearranging the furniture in your house without your knowledge. Before, by scrolling through the feed, you were moving back in time. But suddenly, a post from two days ago appeared at the top of your feed.

2016 年初，Twitter 也不再按时间顺序排列，在用户首次使用时，算法推送信息一度成为默认信息。在应用程序上——对于一个许多人用作实时新闻推送的网站来说，这是一个问题。（按时间顺序排列的选项被称为“Twitter Classic”，仿佛它是一种深受喜爱的垃圾食品口味。）后来，该应用程序会在一段时间后自动将用户切换到算法推送，并强制他们选择退出。尽管Netflix的内容推荐长期以来一直是算法推荐，但2016年也是这家流媒体服务开始改变其主页界面的一年，它优先推荐内容，并根据每个用户进行个性化设置。

Early 2016 was also when Twitter became less chronological, briefly making the algorithmic feed the default when users first got on the app—a problem for a site that many people used as a real-time news ticker. (The chronological option was called “Twitter Classic,” as if it were a beloved junk-food flavor.) Later, the app would swap users over to an algorithmic feed automatically after a while and force them to opt out of it. Although Netflix’s content recommendations had long been algorithmic, 2016 was also when the streaming service began changing its home-page interface, prioritizing recommendations and individualizing it for each user.

这种转变带来了更深刻的文化后果，这或许出乎用户意料，或许也出乎公司本身的意料——就像筑坝改变整个生态系统一样。当信息流被算法化时，它们对不同的人呈现的方式也不同：你不可能知道别人在特定时间看到了什么，因此更难在网上感受到与他人的社群感，那种在电影院看电影或坐下来观看预定的有线电视节目时可能会感受到的集体感。Filterworld 的出现标志着单一文化的瓦解。它有一些优势——我们比以往任何时候都能够消费更广泛的媒体——但它也带来了负面影响。文化应该是社群性的，需要在不同受众之间保持一定程度的一致性；如果没有社群性，它就会失去一些本质的影响力。

Larger cultural consequences, unexpected by users and perhaps by the companies themselves, followed this shift—the way that damming a river changes an entire ecosystem. When feeds are algorithmic, they appear differently to different people: It’s impossible to know what someone else is seeing at a given time, and thus harder to feel a sense of community with others online, the sense of collectivity you might feel when watching a movie in a theater or sitting down for a prescheduled cable TV show. The advent of Filterworld has seen a breakdown in monoculture. It has some advantages—more than ever before, we can all consume a wider possible range of media—but it also has negative consequences. Culture is meant to be communal and requires a certain degree of consistency across audiences; without communality, it loses some of its essential impact.

加剧碎片化问题的原因是，推荐系统更新不会同时推送给所有用户。2016年之后的一两年里，我的Instagram动态仍然严格按照时间顺序排列，而我周围的人都抱怨看不到自己想看的内容。最终，我的动态也转换了，我明白了他们抱怨的原因。我们开始依赖动态的某些运作方式，而当这些方式改变时，我们作为消费者的行为方式也会随之改变。我们被困在算法流程中，被程序设定的变量所驱动。

Intensifying the problem of fragmentation was the fact that recommender-system updates do not roll out at the same time to all users at once across an app. For a year or two after 2016, my personal Instagram feed remained rigorously chronological, while everyone around me complained about not seeing what they wanted. Eventually my feed switched over too, and I understood what they had been complaining about. We came to rely on our feeds working in certain ways, and when those changed, how we behaved as consumers also changed. We were stuck in the algorithmic flow, driven by whichever variables it was programmed to seek.

算法推送的兴起，如同互联网本身一样，缓慢地到来，然后突然爆发。在我撰写本文的2020年代初，推荐系统似乎已成定局，它影响着我们对各种数字媒体的消费。科技常常看似遥不可及，直到开关一触即发，飞跃式的进步变得平凡无奇，成为日常生活中一个再简单不过的事实。

The rise of the algorithmic feed, like the Internet itself, came slowly and then all at once. Early in the 2020s, as I’m writing, recommender systems seem unavoidable, mediating our consumption of every form of digital media. Technology often appears to belong to the distant future right up until the moment the switch flips, and the leap forward becomes totally mundane, a simple fact of daily life.

马塞尔·普鲁斯特在其二十世纪初的长篇小说《追忆似水年华》中，深入挖掘了个人情感在科技发展背景下的微妙变化。在一段文字中，普鲁斯特的叙述者将电话描述为“一种超自然的仪器，它创造的奇迹曾让我们惊叹不已，而如今我们毫不犹豫地使用它，比如召唤裁缝或订购冰淇淋。” 电话直到十九世纪末才被发明，也就是普鲁斯特小说的时代。到1899年，巴黎已有七千名电话用户。然而，电话仍然显得平庸。即使在第一次打电话时，叙述者也对这台设备感到厌烦，而不是敬畏。普鲁斯特写道：“习惯只需要很短的时间来剥离我们与之接触的神圣力量的神秘性，以至于我没有立即接到电话，我立即想到这一切都非常漫长且非常不方便，我几乎决定提出投诉。”

In his sprawling early twentieth-century novel In Search of Lost Time, Marcel Proust excavated such subtle changes in personal sensibilities against the backdrop of evolving technology. In one passage, Proust’s narrator describes the telephone as “a supernatural instrument before whose miracles we used to stand amazed, and which we now employ without giving it a thought, to summon our tailor or to order an ice cream.” The telephone had only been invented in the late nineteenth century, when Proust’s novel is set. By 1899, there were seven thousand telephone subscribers in Paris. And yet telephones had still become banal. Even during one of his first phone calls, the narrator becomes annoyed by the device instead of awed. Proust wrote: “Habits require so short a time to divest of their mystery the sacred forces with which we are in contact, that, not having had my call at once, my immediate thought was that it was all very long and very inconvenient, and I almost decided to lodge a complaint.”

1933年，日本小说家谷崎润一郎以长篇散文《阴翳礼赞》纪念了又一个科技变革时刻。这部散文集讲述了电灯抵达东京的历程。隐喻性的开关已然打开；在谷崎一生中（他生于1886年），由于1867年开始的西方文明的入侵，以及随之而来的全球化浪潮和文化冲突，电灯在日本从默默无闻到无处不在。谷崎写道，西方人“对更明亮光线的追求永无止境”。在散文中，谷崎悼念了日本文化的独特形式，这些形式是由昏暗的烛光所激发的，从家中推拉门上金箔的微光，到昏暗餐厅里味噌汤的朦胧景象：“我们的烹饪依赖于阴影，与黑暗密不可分。”

In 1933, the Japanese novelist Junichiro Tanizaki memorialized another moment of technological change when he wrote In Praise of Shadows, a book-length essay about electric lights arriving in Tokyo. The metaphorical switch had flipped; within Tanizaki’s lifetime (he was born in 1886), electric lights had gone from unknown in his country to ubiquitous, thanks to the intrusion of the West beginning in 1867, in a wave of increasing globalization and subsequent clashes of cultures. The Westerner’s “quest for a brighter light never ceases,” Tanizaki wrote. In the essay, Tanizaki mourned the unique forms of Japanese culture that the old dimness of candlelight had inspired, from the gleam of gold leaf on a home’s interior sliding door to the murky appearance of miso soup in a darkened restaurant: “Our cooking depends upon shadows and is inseparable from darkness.”

然而，谷崎无法忽视电力和其他新设备的吸引力：瓷质马桶、加热器和霓虹灯。“我并非反对现代文明带来的便利，”他写道。正如他在小说中所描述的，这位小说家热爱电影院和现代建筑，就像他欣赏传统一样。《阴翳礼赞》追溯了技术变革、文化适应以及个人品味的转变——这种模式在我们这个时代，《过滤世界》中随处可见。

Yet Tanizaki couldn’t ignore the attraction of electricity and other new devices: porcelain toilets, heaters, and neon signs. “It was not that I objected to the conveniences of modern civilization,” he wrote. As he described in his fiction, the novelist loved movie theaters and modern architecture as much as he appreciated tradition. In Praise of Shadows tracked how technology changed, culture adapted, and personal taste shifted in turn—a pattern we see throughout Filterworld in our own time.

随着新技术的出现，奇迹迅速变得平凡，其功能中的任何小故障都会让人感到烦恼，最终变得可以忽略，奇迹被遗忘。我们忘记了生活并非一直如此，我们无法与远距离的人直接交谈，天花板上的灯并不能照亮每个房间，我们获取的信息和媒体也并非由机器自动过滤。这就是算法信息流如今在我们生活中的存在；算法常常被忽视，成为家具的一部分，只有当它无法像交通信号灯或自来水一样正常运转时才会被注意到。

With new technology, the miraculous quickly becomes mundane, any glitch in its function is felt as bothersome, and finally it becomes ignorable, the miracle forsaken. We forget that life wasn’t always this way, that we couldn’t directly speak to people across long distances, that ceiling lights didn’t make every room bright, or that we didn’t have our information and media automatically filtered by machines. Such is the presence algorithmic feeds now have in our lives; the algorithm is often unconsidered, part of the furniture, noticed only when it doesn’t function in the way it’s supposed to, like traffic lights or running water.

捕捉算法焦虑

CATCHING ALGORITHMIC ANXIETY

如果说会下棋的土耳其机器人是几百年前与神奇科技的（虚幻）邂逅，它能够独立于人类做出决策，那么如今，在我们习以为常的数字空间中，我们每天都会经历数十次这样的体验。机器影响力无处不在，怎么说都不为过。从公开数据来看，Facebook 如今拥有近 30 亿用户，Instagram 约有 20 亿，TikTok 超过 10 亿，Spotify 超过 5 亿，Twitter 有 4 亿，Netflix 超过 2 亿。对于这些平台上的所有用户来说，每一次互动、每一次被动消费，都受到算法推荐的影响。即使有些用户可以选择退出算法推送，他们的参与也会为其他用户的推荐提供数据支持。这种“拉网式”的搜索无可避免。社交网络和流媒体服务已经成为全球相当一部分人口获取信息的主要方式，无论是音乐、娱乐还是艺术。我们如今生活在一个算法文化的时代。

If the chess-playing Mechanical Turk was an (illusionary) encounter with miraculous technology that made decisions independent of a human hand centuries ago, we now undergo that experience dozens of times a day, in the digital spaces that we are accustomed to relying on. It’s hard to overstate the ubiquity of machine influence. From what we can tell using public metrics, Facebook today has nearly three billion users. Instagram has around two billion. TikTok has over one billion. Spotify has over 500 million. Twitter has 400 million. Netflix has over 200 million. For all the people on these platforms, every interaction, every moment of passive consumption, is mediated by algorithmic recommendations. Even if some users can opt out of an algorithmic feed, their participation contributes to the data that fuels other users’ recommendations. The dragnet is inescapable. Social networks and streaming services have become the primary way a significant percentage of the global population metabolizes information, whether it’s music, entertainment, or art. We now live in an era of algorithmic culture.

科技公司长期以来一直致力于实现这种大规模。对这些实体来说，垄断性增长比用户体验的质量更重要，当然也比通过服务信息流公平地传播文化更重要。（数字平台不像艺术博物馆那样承担策展责任。）根据硅谷的理念，追求规模远比它可能带来的任何负面后果更重要，因为马克·扎克伯格在 Facebook 的副手安德鲁·博斯沃思 (Andrew Bosworth) 在 2016 年撰写的一份备忘录表明：

Technology companies have long sought to achieve this massive scale. Monopolistic growth is more important to these entities than the quality of user experience and certainly more important than the equitable distribution of culture through the services’ feeds. (A digital platform has none of the curatorial responsibility of, say, an art museum.) According to Silicon Valley ideology, the pursuit of scale far outweighs any negative consequence it might have, as a memo written by Andrew Bosworth, a deputy of Mark Zuckerberg’s at Facebook, demonstrated in 2016:

所以我们连接了更多人。如果他们把这当成负面因素，那就糟了。也许有人会因为被欺凌而丧命。也许有人会死于利用我们工具策划的恐怖袭击。但我们仍然连接着人们。丑陋的真相是，我们如此坚信连接人与人的重要性，以至于任何能让我们更频繁地连接更多人的事情都*事实上*是好事。

So we connect more people. That can be bad if they make it negative. Maybe it costs someone a life by exposing them to bullies. Maybe someone dies in a terrorist attack coordinated on our tools. And still we connect people. The ugly truth is that we believe in connecting people so deeply that anything that allows us to connect more people more often is *de-facto* good.

这句话鲜明地体现了这样一种态度：只要人们使用一个平台，保持参与度和活跃度，那么无论他们做什么，平台就算成功了。这种持续的参与度依靠自动推荐系统维持，自动推荐系统会推送下一个引人入胜的新闻标题或催眠式娱乐内容。如今，很难想象创造出一种独立于算法推送的文化，因为这些推送控制着它如何呈现给全球数十亿数字消费者。没有推送，就没有受众——创作只存在于其创作者及其直接联系的人心中。更难以想象在算法推送之外消费任何东西，因为它们的推荐不可避免地会影响电视、广播和书籍的内容，即使这些体验并不包含在推送中。过滤世界无处不在。

That statement is a stark illustration of the attitude that if people are using a platform, staying engaged and active, then it counts as successful—no matter what they are doing. That ongoing engagement is sustained by automated recommendations, delivering the next provocative news headline or hypnotic entertainment release. Today, it is difficult to think of creating a piece of culture that is separate from algorithmic feeds, because those feeds control how it will be exposed to billions of consumers in the international digital audience. Without the feeds, there is no audience—the creation would exist only for its creator and their direct connections. And it is even more difficult to think of consuming something outside of algorithmic feeds, because their recommendations inevitably influence what is shown on television, played on the radio, and published in books, even if those experiences are not contained within feeds. Filterworld spills out everywhere.

特雷弗·博芬是一位戏剧学者，曾担任高中教师。他对算法文化的本质做了一个贴切的描述：“票房好的电影都是在TikTok上拥有粉丝的电影；Billboard百强单曲榜由TikTok决定；你去巴诺书店，会看到一张BookTok的桌子，”他说。（BookTok是TikTok文学影响者社区的术语。）换句话说，一种文化要想在商业上取得成功，它必须已经在数字平台上拥有了一定的影响力。博芬的职业生涯也受到了算法推送的影响。当他开始和十几岁的学生一起学习TikTok舞蹈动作，并把视频发布到网上时，他很快就积累了数十万的粉丝。Instagram 和其他平台上。他出现在国家电视台，并短暂地成为了一个热门人物——舞蹈老师。根据他的经历，他出版了一本关于舞蹈表演的学术专著。随着舞蹈表演在TikTok上日益普及，这一主题迅速引起了大学和编辑们的关注。“今年一个月里，人们对我作品的兴趣比过去十年加起来还要多，”博福内告诉我。

Trevor Boffone, a scholar of theater who took up work as a high school teacher, gave me an apt description of what algorithmic culture amounts to: “The films that do well are films that have TikTok followings; the Billboard Hot 100 is dictated by TikTok; you go to Barnes and Noble and you see a BookTok table,” he said. (BookTok is a term for TikTok’s community of literary influencers.) In other words, for a piece of culture to be commercially successful, it must already have traction on digital platforms. Boffone’s career, too, has been shaped by algorithmic feeds. When he began learning TikTok dance moves with his teenage students and posting videos of them online, he quickly accrued hundreds of thousands of followers on Instagram and other platforms. He appeared on national television, briefly becoming a viral character—the dancing teacher. Following his experiences, he published an academic monograph on dance performance, a subject that had quickly become more compelling to universities and editors with its rising public popularity on TikTok. “I’ve had more interest in one month of this year in my work than in the previous ten years combined,” Boffone told me.

Boffone 的经历遵循了 Filterworld 的一个基本规则：在算法推送下，热门内容会更受欢迎，冷门内容则会更加少见。成功或失败都会加速发生。“一篇传统的 Instagram 帖子，其生命力取决于帖子的前三到五分钟，”Boffone 说。如果帖子能立即获得互动，那么它就更有可能获得更多，反之亦然。这种动态可能很残酷。当我发布一张另类的 Instagram 图片或一条冷门推文，却没有获得太多关注时，即使我知道自己并没有中算法大奖，我也会多次查看帖子，希望能获得更多点赞。

Boffone’s experience follows a fundamental rule of Filterworld: Under algorithmic feeds, the popular becomes more popular, and the obscure becomes even less visible. Success or failure is accelerated. “A traditional Instagram post, the life of it is dictated by the first three to five minutes of the post,” Boffone said. If a post gets engagement immediately, then it’s likely to get more, and vice versa. This dynamic can be cruel. When I post an offbeat Instagram image or an obscure tweet and it doesn’t get much action, that doesn’t stop me from checking back multiple times for more likes, even though I know I haven’t hit the algorithmic jackpot.

缺乏关注不可避免地引发了一个问题：信息流将推广什么？它暗中鼓励更安全的选择，敦促人们趋同。谁获得推广也是一个问题。通常，获得赞誉、关注，从而从算法信息流中的流行中获得经济利益的，并不是模因或潮流的原创者。TikTok 编舞就是一个例子。TikTok 网红 Charli D'Amelio 于 2019 年因其在该平台上的舞蹈视频而出名。但她推广并经常被认为是她的一个动作，名为 Renegade，实际上是早些时候由来自佐治亚州的黑人青少年 Jalaiah Harmon 创作的。Renegade 是一系列非常适合 TikTok 屏幕的正面动作，包括挥拳和扭臀——这个动作并不太难，但也很难记住，因此重复表演会很有成就感。

The absence of attention inevitably raises the question of what the feed will promote, tacitly encouraging safer choices, urging conformity. Who receives promotion is also a problem. It’s often not the original creators of a meme or trend who get credit, attention, and thus financial gain from its popularity in an algorithmic feed. TikTok choreography itself is an example. The TikTok influencer Charli D’Amelio became famous in 2019 for her dance videos on the platform. But one of the moves she popularized and was often credited with, called the Renegade, was actually created earlier by Jalaiah Harmon, a Black teenager from Georgia. The Renegade was a series of front-facing movements perfect for the TikTok screen, with swinging punches and hip shakes—not too difficult a sequence, but also tough to memorize and thus rewarding to re-perform.

哈蒙最初在一款名为 Funimate 的应用程序和 Instagram 上发布了这支舞蹈。但 TikTok 的超级算法推送加速了这支舞蹈的主流知名度，这得益于达梅利奥的粉丝，尽管由于达梅利奥没有引用哈蒙的名言，它也抹去了哈蒙的作者身份。来自边缘群体的内容创作者不像达梅利奥这样的白人、私立学校毕业、受过专业训练的舞蹈家那样，能够接触到媒体和获得关注，因此他们拥有很难从 Filterworld 的潮流中获益。（自从 Harmon 的工作获得认可以来，她本人的 TikTok 粉丝也增加了 300 万。）

Harmon first posted the dance on an app called Funimate as well as Instagram. But TikTok’s hyper-algorithmic feed accelerated it to mainstream fame, seeded by D’Amelio’s following, even as it helped to erase Harmon’s authorship, since D’Amelio didn’t cite her. Content creators from marginalized groups, who don’t have the same access to media and attention as, say, a white, private-school-educated, professionally trained dancer, like D’Amelio, have a harder time benefiting from the tides of Filterworld. (Since being recognized for her work, Harmon herself has gained three million TikTok followers.)

鉴于这些反复无常的系统控制着我们生活的方方面面，从与朋友的社交到为我们的创意项目吸引受众，社交媒体用户感到偏执也就不足为奇了。我们被鼓励忽略算法流程，但它们的故障却提醒我们，它们拥有不劳而获的权威。算法影响的模糊性创造了一种被称为“算法焦虑”的感觉。算法焦虑描述了一种日益增长的意识，即我们必须不断应对我们无法理解和控制的自动化技术流程，无论是在我们的Facebook动态、谷歌地图驾车路线，还是亚马逊产品促销中。我们总是在预测和猜测算法做出的决策。算法焦虑不是假设或抽象的概念：它已经普遍存在。科技公司意识到了这一点，并且多年来一直在操纵用户的这种感受。

Given that these capricious systems control so many facets of our lives, from socializing with our friends to building audiences for our creative projects, is it any wonder that social media users feel paranoid? We’re encouraged to overlook algorithmic processes, but their glitches remind us of their unearned authority. The ambiguity of algorithmic influence creates a feeling that has been labeled “algorithmic anxiety.” Algorithmic anxiety describes the burgeoning awareness that we must constantly contend with automated technological processes beyond our understanding and control, whether in our Facebook feeds, Google Maps driving directions, or Amazon product promotions. We are forever anticipating and second-guessing the decisions that algorithms make. Algorithmic anxiety is not hypothetical or an abstraction: It’s already prevalent. Technology companies are aware of it and have been manipulating the feeling in their users for years.

2018年，时任佐治亚理工学院博士生的Shagun Jhaver与两名Airbnb员工合作，对公司用户进行了一项社会学研究。他们分析了平台的房东（通过Airbnb出租房屋获取收入）如何与Airbnb的算法推荐、搜索和评分系统互动，以及他们对该系统的感受。这些系统帮助租客找到并预订房源。团队在研究结果中写道，Jhaver和其他研究人员创造了“算法焦虑”一词，指的是房东“对Airbnb算法运作方式的不确定性以及感知到的缺乏控制力”。房东担心搜索算法不公平地忽略他们，或优先考虑其他房源。Jhaver注意到，这种焦虑更多地归因于技术，而非房东出租房屋的实际质量：“这主要与算法本身有关，而不是通过其他方式改进他们的房源和房产，”他告诉我。

In 2018, Shagun Jhaver, at the time a PhD candidate at the Georgia Institute of Technology, worked with two Airbnb employees to conduct a sociological study of the company’s users. They analyzed how the platform’s hosts—who rented out their homes on the service for income—interacted with and felt about Airbnb’s algorithmic recommendation, search, and ratings systems, which helped renters find and book their listings. Jhaver and the other researchers coined the term algorithmic anxiety for the hosts’ “uncertainty about how Airbnb algorithms work and a perceived lack of control,” the team wrote in their findings. Hosts worried that the search algorithm was unfairly ignoring them or prioritizing other properties. Jhaver noticed that the anxiety was ascribed more to the technology than the quality of the actual homes the hosts were renting out: “It was particularly to do with the algorithm itself rather than improving their listing and property in other ways,” he told me.

研究人员写道，Airbnb 迫使房东进行“双重谈判”，因为他们必须确定房客在房源中寻找什么，以及算法优先考虑哪些变量来更频繁地推广他们的房源。但房东可以他们无法判断哪些变量真正提升了他们的房源排名。他们认为，诸如累积评论数量、评论质量以及可用照片数量等因素有助于提升他们的排名，但他们不太确定算法是否分析了他们的定价、房屋设施或房东的租期。他们对系统如何运作知之甚少。这完全取决于个人认知。正如研究中一位房东抱怨的那样：“看到搜索结果，真是令人沮丧：很多比我的房源更差的房源排名却更高。”

Airbnb forces a “double negotiation” for the hosts, the researchers wrote, because they must determine what their guests are looking for in a listing as well as which variables the algorithms are prioritizing to promote their property more often. But hosts could not tell which variables actually boosted their listing. They believed factors like the number of reviews accrued, the quality of reviews, and the number of photos available would help their chances of promotion, but they were less certain as to whether the algorithm analyzed their pricing, home amenities, or length of tenure as a host. They had little information about how the systems worked. It was all a matter of perception. As one host in the study complained: “It’s frustrating seeing the search: lots of listings that are worse than mine are in higher positions.”

当然，质量是主观的，但主持人的感受恰恰说明了用户可能会因为算法评估而感到被误解和误判。“这就像一场考试，但你不知道考试内容是什么，也不知道如何才能取得好成绩，”Jhaver 解释道。而且，不只是用户不知道发生了什么。Jhaver 继续说道：“归根结底，即使是算法的开发者也无法告诉你哪个因素导致了哪个决策；算法的复杂性如此之高，以至于无法区分不同的因素。”

Quality is subjective, of course, but the host’s sentiment speaks to how users can feel misunderstood and misjudged by algorithmic evaluations. “It’s like an exam, but you don’t know what’s going to be on this exam, or how to score well on this exam,” Jhaver explained. And it’s not just the users who don’t know what’s going on. Jhaver continued: “At the end of the day, even the people who create the algorithms cannot tell you which factor was responsible for which decision; the complexity of the algorithm is so high that disentangling different factors is just not possible.”

无法与算法博弈可能会导致房东收入立即下降，而他们和其他劳动者一样，都依赖于收入的稳定。（算法推广的不一致性迫使我们不得不与之互动，并为此付出更多努力，就像反复拉动老虎机的杠杆以赢得大奖一样。）像 Airbnb 这样的零工经济平台长期以来一直承诺提供灵活的工作和其他谋生或补充收入的方式，但它们也创造了一种新的劳动形式，即需要随时了解算法优先级的变化。房东担心 Airbnb 的搜索算法，艺术家担心 Instagram 的算法，音乐家担心 Spotify 的算法。研究人员发现，房东对这种算法焦虑的反应是发展“民间理论”——一些迷信的伎俩，旨在获得更多的算法推广和更好的搜索结果——就像我过去用虚假的结婚公告发布文章链接一样。这些策略包括不断更新房源日历、更改个人资料详细信息，甚至在一天中更频繁地打开 Airbnb 网站。这些伎俩让人想起一个孩子把勺子放在枕头下制造雪花，或许同样有效。因为研究人员发现，主持人“通常会怀疑这些理论是否正确，但尽管不确定，他们仍然会执行这些操作以试图影响算法。”

Failing to game the algorithm may cause an immediate drop in income for hosts, which they, as any worker, rely on remaining consistent. (The inconsistency of algorithmic promotion forces us to engage with it and stress about it even more, like repeatedly pulling a slot machine lever to hit the jackpot.) Gig-economy platforms like Airbnb have long promised flexible work and alternative ways of making or supplementing a living, but they also created a new form of labor in the need to stay up to date on changes in algorithmic priorities. Where hosts worry about Airbnb’s search algorithm, artists similarly fret about Instagram’s and musicians about Spotify’s. The hosts’ reaction to such algorithmic anxiety, the researchers found, was to develop “folk theories”—superstitious tricks that were meant to goose more algorithmic promotion and better search results—the same way I used to post my article links with fake wedding announcements. Some of the strategies included constantly updating their listings calendar, changing their profile details, and even opening the Airbnb website more often throughout the day. The tricks bring to mind a child putting a spoon under their pillow to cause a snow day and are perhaps equally as effective. As the researchers found, hosts “usually had doubts about whether such theories were true but despite their uncertainty still performed those actions in an attempt to influence the algorithm.”

算法焦虑可以说是一种当代瘟疫。它在许多用户中引发了类似强迫症的倾向，使他们过度警觉，并需要重复同样的仪式。因为当这些仪式“奏效”时，效果非常显著，不仅会因获得关注而产生心理多巴胺激增，而且如果你的网络形象被货币化，还能获得潜在的经济回报。它影响着我们许多在线行为：选择合适的头像，在 Instagram 账户上精心挑选一组吸引人的照片，在市场列表中选择合适的关键词。我们担心自己的帖子要么不会被合适的人看到，要么担心如果被选中进行病毒式传播，会被太多人看到，从而将我们暴露给陌生人。这种对关注的追求会带来情感上的后果：我们最终会变得既兴奋过度又麻木不仁，就像一个目光呆滞、等待匹配符号出现的老虎机玩家。

Algorithmic anxiety is something of a contemporary plague. It induces an OCD-ish tendency in many users toward hyperawareness and the need to repeat the same rituals, because when these rituals “work,” the effect is so compelling, resulting in both a psychological dopamine rush from receiving attention and a potential economic reward if your online presence is monetized. It undergirds so many of our behaviors online: selecting the right profile picture, curating an attractive grid of photos on an Instagram account, choosing the right keywords on a marketplace listing. We worry that our posts either won’t be seen by the right people or will be seen by too many if selected for virality, exposing us to strangers. There’s an emotional fallout to this quest for attention: we end up both overstimulated and numb, much like a glassy-eyed slots player waiting for matching symbols to come up.

算法焦虑的产生是因为用户和算法之间存在着巨大的不对称关系。对于个人用户而言，试图预测或决定算法的结果就像试图控制潮汐。继续这个比喻，用户所能做的就是在已经形成的浪潮中冲浪。公司几乎没有动力去缓解这种焦虑，因为用户的困惑可能对业务有利。当一家公司的产品无效或用户遇到困难时，可以将其归咎于“算法”这个不透明的实体，它被视为用户和公司本身的外部实体，因为它们被比作不透明的“黑匣子”。对算法的利用被伪装成意外故障，而不是有意为之的公司政策。实际上，像 Facebook 这样的公司完全控制着他们的算法系统，能够随意更改它们，或者关闭它们。

Algorithmic anxiety happens because there is a dramatically asymmetrical relationship between user and algorithm. For the individual user, trying to predict or dictate the outcome of the algorithm is like trying to control the tide. To continue the metaphor, all users can do is surf the wave that’s already formed. There is little incentive for companies to assuage this anxiety because a user’s confusion can be beneficial to business. When a company’s product is ineffective or a user encounters difficulty, it can be blamed on the opaque entity of “the algorithm,” which is perceived as external to both the users and the company itself, since they are likened to opaque “black boxes.” Exploitation is disguised as an accidental glitch instead of an intentional corporate policy. In reality, a company like Facebook is wholly in control of their algorithmic systems, able to change them at will—or turn them off.

算法焦虑将行动的负担放在个人身上，而不是企业身上——用户必须改变自己的行为，否则就会面临消失的风险。用户有时会抱怨自己在平台上的帖子或内容突然缺乏之前的互动度，从而被“影子封禁”。用户常常担心自己的账号被某些人专门封禁，却没有任何预警或追索权。决策者；但算法的优先级可能只是悄无声息地发生了变化，流量不再流向它们。这种效应可以追溯到“土耳其机器人”（Mechanical Turk）；我们无法总是区分技术在发挥作用还是技术在发挥作用的幻觉，但最终，感知可能与现实一样具有影响力。

Algorithmic anxiety places the burden of action on the individual, not the business—the user must change their behavior or risk disappearing. Users sometimes complain of being “shadowbanned” when their posts or content on a platform suddenly lack the same level of engagement as before. Users often fear that their account specifically has been blocked without warning or recourse by some decision-maker; but the algorithmic priorities may simply have silently changed, and traffic is no longer flowing in their direction. The effect goes back to the Mechanical Turk; we can’t always tell the difference between technology working and the illusion of technology working, but the perception may be just as impactful, in the end, as the reality.

学者帕特里夏·德·弗里斯在其2019年的论文《当代艺术中的算法焦虑》中将算法焦虑定义为一种“自我被认为受到算法机制的限制、约束和支配”的状态。她的措辞令人叹为观止。我们自身感知到的可能性——我们的表达和创作模式——如今存在于数字平台的结构之中。德·弗里斯写道，这种焦虑的后果包括“算法决定论、宿命论、犬儒主义和虚无主义”。它构建了一种感觉：既然我们用户无法控制技术，我们不妨屈服于算法文化的局限，并将其视为不可避免的结果。许多用户已经陷入了这种绝望的状态，既感到不满，又无法想象其他的替代方案。

In her 2019 dissertation titled Algorithmic Anxiety in Contemporary Art, the scholar Patricia de Vries defined algorithmic anxiety as a condition in which “the possible self is perceived to be circumscribed, bounded, and governed by algorithmic regimes.” Her words feel breathtakingly accurate. The possibilities that we perceive for ourselves—our modes of expression and creation—now exist within the structures of digital platforms. The consequences of such anxiety include “algorithmic determinism, fatalism, cynicism, and nihilism,” de Vries wrote. It builds to a sense that, since we users cannot control the technology, we may as well succumb to the limits of algorithmic culture and view it as inevitable. Many users have already entered such a state of despair, both dissatisfied and unable to imagine an alternative.

早在 2013 年，德弗里斯就开始观察这种文化变迁，当时她看到几场博物馆展览重点展出了批判自动监控和数据收集的艺术家的作品。尽管算法信息流才刚刚开始进入主流体验，但诸如 2010 年由算法股票交易引发的“闪电崩盘”以及面部识别等技术，已将“算法”一词植入新闻头条。到了 2010 年代中期，当她开展研究时，它已成为“我们着迷的对象”，德弗里斯告诉我。算法是一个幽灵，萦绕在我们与数字平台的任何接触中，以及它们在我们生活中日益侵入的存在中。这并不是说我们理解算法本身到底在做什么：“正如恐高症与高度无关一样，算法焦虑也不仅仅与算法有关，”德弗里斯说。

De Vries began observing this cultural shift as early as 2013, when she saw several museum exhibitions highlighting the work of artists who were critical of automated surveillance and data collection. While algorithmic feeds had only just begun entering mainstream experience, events like the 2010 Flash Crash, caused by algorithmic stock trading, and technology like facial recognition had implanted the word in news headlines. By the middle of the decade, when she developed her research, it became “this sort of object of our fascination,” de Vries told me. The algorithm was a specter that haunted any encounter with digital platforms and their increasingly intrusive presence in our lives. That is not to say we understood what exactly algorithms were doing, per se: “Just as the fear of heights is not about heights, algorithmic anxiety is not simply about algorithms,” de Vries said.

为了向前迈进，我们必须将算法推荐作为一种技术的影响，与我们习惯性地将其作为在线交流的主要守门人的方式区分开来。毕竟，算法与数据密不可分。它们不断运转，而这些机制是由人类创造并不断更新的。实际的影响力与对影响力的恐惧并存，后者同样具有操纵性。算法通过承诺为我们做决定、预测我们的想法和愿望而进入了我们的生活。“过滤世界”代表了算法精神世界的建立——不仅仅是它们如何运作，还包括我们用户如何依赖它们，允许它们取代我们自身的自主权，即使我们开始憎恨它们迫在眉睫的存在。

To move forward, we must disentangle the effects of algorithmic recommendations as technology from the ways that we have habitually adopted them as the primary gatekeepers of our online communication. Algorithms, after all, are inextricable from the data they run on, which has been created and is constantly refreshed by humans. Actual influence coexists with the fear of influence, which is equally manipulative. Algorithms entered our lives by promising to make decisions for us, to anticipate our thoughts and desires. Filterworld represents the establishment of the psychic world of algorithms—not just how they work, but how we users have come to rely on them, allowing them to displace our own agency, even as we come to resent their looming presence.

亚马逊书店

THE AMAZON BOOKSTORE

一天下午，在华盛顿特区，我向东南方向步行前往乔治城，这里是这座城市的主要购物中心之一。鹅卵石街道和河畔主干道构成了一个类似户外购物中心的购物区，可爱的餐厅和纸杯蛋糕烘焙坊散布在耐克、露露柠檬、Zara、Club Monaco 等国际时尚品牌的零售店之间。购物者会被最适合自己的品牌所吸引，或许是基于他们独特的审美观，或许是源于一种隐含的生活方式。顾客选择的品牌不仅反映了品牌本身，也反映了顾客自身的价值——这是一种创造者和顾客之间的互惠关系，顾客相信（或希望）品牌的理念能够代表他们自身的身份认同。顾客去耐克不仅仅是因为他们想要一双好的跑鞋；耐克为其生产的所有产品都注入了青春和活力的气息，无论是运动鞋还是色彩鲜艳的丝网印花T恤。 Zara 不仅生产连衣裙，还生产其他众多服装单品，并通过这些单品展现出一种前卫的时尚感，而价格却不像传统奢侈品牌那么昂贵。购买 Zara 连衣裙，就等于参与其高效魅力的形象——一场彻夜派对。店里的每一件单品都体现了这种感觉。换句话说，每家店都服务于一种特定的感觉，消费者可以根据自己的喜好进行选择。

One afternoon, in Washington, D.C., I walked southeast to Georgetown, which is one of the city’s main shopping hubs. Its cobblestoned streets and main thoroughfare by the river form something like an outdoor mall, with cutesy restaurants and cupcake bakeries sprinkled among the retail outlets of international fashion brands: Nike, Lululemon, Zara, Club Monaco. Shoppers gravitate to whichever brands suit them best, perhaps based on a particular aesthetic or an implied lifestyle. The brand a customer picks says as much about them as it does about the brand—it’s a reciprocal relationship between creator and customer, with the customer believing (or hoping) that the brand’s ethos is representative of their own identity. A shopper doesn’t go to Nike just because they want good running shoes; Nike casts an aura of youth and energy over everything it produces, whether sneakers or brightly screen-printed T-shirts. Zara makes dresses, among many other clothing items, and through them projects a sense of cutting-edge style without the cost of traditional luxury brands. To buy a Zara dress is to participate in its image of efficient glamour—a nightly party. Every piece in the store refers to that sensibility. In other words, each store serves a particular sensibility, and consumers select from them based on their individual tastes.

但乔治敦有一家书店却与众不同：亚马逊书店，这家大型互联网公司运营的实体书店。亚马逊于 2015 年开始建立实体书店，第一家店位于西雅图。当它于 2018 年在华盛顿特区开业时，在现实世界中偶然发现它的无衬线字体标识是很奇怪的（那时它的卡车还没有出现在城市街道和高速公路上）。书店的内部装饰也很陌生。当我第一次走进去时，我被视觉上的混乱所震惊——它没有书店通常那种冥想式的平静氛围。相反，它更像一家杂货店。所有的书封面都朝外，像许多数字图标一样并排摆放在书架上。每本书下面都有一个数字标签，显示其在亚马逊网站上的相应排名，排名由用户评论、销售量甚至购买者阅读的页数决定，这一指标由亚马逊的 Kindle 电子阅读器以数字方式衡量。

But one Georgetown store stood out as different: Amazon Books, a brick-and-mortar bookstore operated by the enormous Internet corporation. Amazon began building physical bookstores in 2015, with its first location in Seattle. When it opened in D.C. in 2018, it was strange to stumble across its sans serif logo out in the real world (this was before its trucks became omnipresent on city streets and highways). The bookstore’s interior was also unfamiliar. When I walked in for the first time, I was struck by the visual chaos—it didn’t have a bookstore’s usual atmosphere of meditative calm. Instead, it more closely resembled a grocery store. All the book covers were facing outward, side by side on the shelf like so many digital icons. And each book had a digital label below it showing its respective rank on the Amazon site, with numbers determined by user reviews, volume of sales, and even how many pages buyers of the book read through, a metric measured digitally by Amazon’s Kindle e-readers.

实体书店的布局沿用了其网站的设计，“最受欢迎”的书目占据了显著位置。书籍的排列并非按照作者、作者国籍，甚至也不是一贯按照类型排列，而是根据它们在网上的成功程度——这是亚马逊图书用来衡量文学质量和价值的算法。互动再次占据了主导地位。书店周围张贴的标牌解释了某些书籍被展示的原因：它们是“畅销书”，或者被评为“4.5 星及以上”，甚至“4.8 星及以上”。平均多出的 0.3 星真的能说明一本书的价值吗？其他书目则是“亚马逊网站上最受欢迎”的书目，或者根据它们在网上收到的预订数量进行突出显示。一些书架墙被布置成一个模拟推荐系统。每个书架的左侧是一本书，标签为“如果你喜欢← ”，右侧则是一系列书籍，标签为“你会喜欢→ ”。例如，诺亚·哈拉里的《人类简史》引发了其他非虚构类畅销书的推荐，例如贾雷德·戴蒙德的《枪炮、病菌与钢铁》和彼得·弗兰科潘的《丝绸之路》。然而，最引人注目的是，每本书的定价都根据亚马逊网站的算法进行，该算法会根据供求情况实时调整价格。

The arrangement of the physical store followed the design of its website, with the “most popular” titles featured prominently up front. The books were not organized by author, their author’s nationality, or even consistently by genre, but rather by how successful they were online—that was the algorithm Amazon Books used for determining the quality and value of literature. Engagement reigned supreme once again. Signs posted around the store explained why certain books were shown off: they were “top sellers,” or they were rated “4.5 Stars & Above,” or even “4.8 Stars & Above.” Did the extra 0.3 stars on average really indicate that much about a book’s worthiness? Other titles were “Most-Wished-For on Amazon.com” or else highlighted by how many preorders they had received online. Some walls of shelves were arranged into an analog recommendation system. On the left side of each shelf was a single book, labeled “If You Like ←,” with a selection of books to the right, “You’ll Love →.” For example, Noah Harari’s Sapiens spurred a recommendation of other nonfiction bestsellers like Jared Diamond’s Guns, Germs, and Steel and Peter Frankopan’s The Silk Roads. Most remarkable of all, however, was that each book was priced according to Amazon’s website algorithm, which adjusts prices in real time based on supply and demand.

这与独立书店截然相反，后者以其魅力和独特的个性而闻名。它们的书架标签独特或另类，展示着一些特色作品：新时代、艺术专著、地方史。店主了解当地客户，巧妙地调整库存以反映当地客户群。有些书可能卖得不太好，但店主选择将它们摆放展示，以展现书店的理念和个人品味。

This was the opposite of independent bookstores, which have a well-deserved reputation for charm and personal quirkiness. Their shelf labels are unique or offbeat, presenting specialties: New Age, art monographs, local history. Their owners know their local clientele, subtly shaping the stock to reflect it. Some books may not sell particularly well, but the owners choose to keep them on display as a demonstration of the store’s ideals and their personal taste.

在亚马逊书店里很​​难感受到任何品味；那里没有我能认同的精神。相反，整体叙事完全由市场和任何能引起注意的东西驱动。所展现的美学风格与肤浅、即时的互动需求紧密相连，这或许是2010年代后期小说封面流行的部分原因，封面由色彩鲜艳、半抽象、半具象的斑点组成。（这些斑点或许很吸引人，但与小说内容无关。）虽然亚马逊书店里有很多和我经常光顾的其他书店一样的书籍，但店内环境却让人难以辨认，感到疏离。我无法把自己视为一个买家，这不仅是因为亚马逊作为雇主和垄断企业的破坏性做法，还因为店里缺乏对理想购物者的清晰概念。

It was hard to discern a sense of taste in Amazon Books; there was no spirit there with which I could identify. Instead, the overall narrative was driven wholly by the market and whatever provoked attention. The aesthetics on display were yoked to the need for superficial, immediate engagement, perhaps part of the reason for the late-2010s vogue for novel covers made up of brightly colored, semi-abstract, semi-figurative blobs. (They may have been eye-catching, but had nothing to do with the novels’ content.) Though the Amazon store stocked many of the same books as other bookshops I frequented, the environment was unrecognizable and alienating. I couldn’t see myself as a buyer, not only because of Amazon’s destructive practices as an employer and as a monopolistic corporation, but because there was no coherent idea of an imagined shopper.

就像互联网规模下运营的许多事物一样，书店缺乏人性。它根据所有亚马逊客户产生的海量数据进行推荐。就好像你只能购买《纽约时报》畅销书排行榜上的书籍，但这份排行榜却由一家不值得信赖的公司操控，这家公司一心想把书籍当作可替代的物品，以便尽快甩卖。亚马逊以销量衡量图书，就像Facebook以点赞数评估内容一样，这种做法与我理想中的书店——一个逃离现实、发现新鲜事物的地方——截然相反。

Like many things that operate at the scale of the Internet, the bookstore was inhuman. It made recommendations based on the mass of data produced by all Amazon customers. It was as if you could buy only the books that appeared on the New York Times bestseller list, but the list was operated by an untrustworthy company, one solely devoted to treating books as fungible objects to be offloaded as quickly as possible. Amazon measured books by sales the way Facebook evaluates content by likes, an approach antithetical to my romantic vision of a bookstore as a place to escape, to discover something new and surprising.

亚马逊书店之所以让人感觉如此怪异，原因之一是它代表着互联网算法逻辑对我们所谓的“现实生活”的公然入侵。我们被网络上的歌曲、图片和电视节目推荐轰炸；当这些体验发生在屏幕上时，我们很容易接受这些推荐和算法的自动调节，因为这些建议是如此顺畅、快速且不可避免。但当它们发生在现实世界中时，它们就更具干扰性，因为我们的选择很少如此明显地由机器决定。或许亚马逊书店的怪异之处源于它让我直面自身缺乏自由，表明算法在多大程度上迫使我们不为自己思考。

One reason the Amazon bookstore felt so strange was that it represented a blatant intrusion of the Internet’s algorithmic logic into what we call “real life.” We’re bombarded by suggestions for songs, images, and television shows online; it’s easy to accept the recommendations and the automatic mediation of algorithms when the experience happens on a screen, because these suggestions are so frictionless, fast, and unavoidable. But they are much more obtrusive when they take place in the physical world, where our choices are rarely dictated so obviously by machines. Maybe the Amazon store’s uncanniness stemmed from how it confronted me with my own lack of freedom, demonstrating just how much algorithms push us into not thinking for ourselves.

亚马逊所有数据平均值驱动的书店选择出奇地同质化，最终也乏味无趣。为了吸引我——或者至少吸引我这个普通消费者——书店事先经过了严格的筛选，并且用大量保证保证其他人也喜欢这些书。但我并没有感到兴奋，也没有受到鼓励去翻阅任何一本书。相反，我感到不知所措，这或许是过滤世界中消费者的默认状态：被过剩的内容包围，却没有任何灵感。

The bookstore selection driven by the average of all of Amazon’s data was curiously homogenous and ultimately boring. It had been aggressively filtered in advance to appeal to me—or at least myself as a generic consumer—with abundant reassurances that other people did like the books on display. But I wasn’t excited or encouraged to page through any of them. Rather, I was overwhelmed, which might be the default state of consumers in Filterworld: surrounded by superabundant content, but inspired by none of it.

谷歌工程师切特·哈塞（Chet Haase）2017年在推特上写的一则笑话直指问题所在：“一个机器学习算法走进一家酒吧。酒保问：‘你想喝点什么？’算法说：‘其他人都喝点什么？’” 笑话的妙处在于，在算法文化中，正确的选择永远是大多数人已经做出的选择。但即使其他人都做出了正确的选择，你或许只是没心情喝威士忌酸酒而已。

A joke written on Twitter by a Google engineer named Chet Haase in 2017 pinpoints the problem: “A machine learning algorithm walks into a bar. The bartender asks, ‘What’ll you have?’ The algorithm says, ‘What’s everyone else having?’ ” The punch line is that in algorithmic culture, the right choice is always what the majority of other people have already chosen. But even if everyone else was, maybe you’re just not in the mood for a whiskey sour.

我最喜欢的书店一直是麦克纳利·杰克逊书店 (McNally Jackson)，它是由纽约市区多家书店组成的连锁店，最初开在苏荷区（2023年，原旗舰店迁至六个街区之外）。虽然它本身并没有什么个性化的元素，但总感觉这家书店就是为我而设，仿佛它了解我的需求，因为它的藏书既丰富又精准。在布鲁克林生活的十年里，我经常乘坐地铁L线到曼哈顿，再乘坐地铁6号线到市中心，沿着绿树成荫的街道走到麦克纳利店面，穿过玻璃门进入前厅，最终进入宽敞的店铺。店面前面两张大小相同的桌子构成了它的核心：左边是非虚构类书籍，右边是虚构类书籍，后面是相应的书架。陈列的书籍中有很多是近期出版的，但也有一些是店员挑选出来的值得仔细阅读的书籍，这是一种集体的精心策划。一本小型出版社的哲学学术著作，紧挨着畅销非虚构类书籍；小说类书籍的陈列台上不仅摆放着小说，还有诗歌、杂志、虚构回忆录以及混合体裁的书籍。这样的摆放方式仿佛在告诉读者：相信我们就好。

My favorite bookstore has long been McNally Jackson, a collection of New York City–area stores whose first location was in SoHo (in 2023, the original flagship relocated six blocks away). Though it wasn’t in any way personalized, it always felt like the store was there for me personally, like it understood what I was looking for, because its stock was both so broad and so specific. During my decade living in Brooklyn, I often took the L train into Manhattan and then the 6 downtown, walking the leafy streets to the McNally storefront, through the glass door into the vestibule, and then into the contained sprawl of the shop. Two equally sized tables in front formed its nucleus: on the left, a table for nonfiction, and on the right, fiction, with corresponding sets of shelves behind them. Many of the books stacked on display were recent releases, but there was also a selection of the books the store’s staff decided were worth a closer look, a collective act of curation. An academic work of philosophy from a small press was nestled next to popular nonfiction; the fiction table didn’t just hold novels but poetry, zines, fictionalized memoirs, and hybrid-genre books. The arrangements always seemed to say: Just trust us.

McNally 表的内容每周都会发生变化。每次重新排列的背后，我都能感受到挑选书籍的手，这是一种个人智慧，而非单一的公式。浏览是发现新事物的一种方式；有人可能会说，亚马逊的“如果你喜欢这个，你也会喜欢那个”的公式功能类似，但麦克纳利书店的联系不那么直接，也不那么字面化。它们拓展了购物者对特定品类所能包含内容的理解。

The contents of the McNally tables shifted on a weekly basis. Behind each rearrangement I felt the hand that selected the book, an individual intelligence, rather than a singular formula. Browsing was a way of discovering new things; one could argue that Amazon’s formula of “if you like this, you’ll like that” functions similarly, but the connections at McNally were less direct and literal. They expanded the shopper’s idea of what a particular category could contain.

如果说亚马逊书店代表着算法逻辑的胜利，那么麦克纳利就是人类潮流引领者的巅峰——我们常常用这个词来形容那些对我们消费的文化进行分类和筛选的人。书商是潮流引领者，但为顾客推荐书籍的图书管理员、生活方式精品店的专业买手、电台DJ、代表电影在全国影院上映的电影预订代理人，以及为演出场地预订乐队的音乐会策划人也同样如此。这些潮流引领者都在文化创造者和消费者之间搭建了沟通的桥梁。他们不断收集和评判新的素材，以确定它如何以及为何能够引起观众的共鸣——这一过程如今被统称为“策展”。

If the Amazon bookstore represented the triumph of algorithmic logic, then McNally was the pinnacle of human tastemakers, the word we often use for the people who sort and select the culture that we consume. Booksellers are tastemakers, but so are librarians who recommend titles for their patrons, professional buyers for lifestyle boutiques, radio-station DJs, movie booking agents who advocate on behalf of films to theaters nationwide, and concert programmers who book bands for venues. These tastemakers all provide an interface between the creators of culture and its consumers. They constantly gather and judge new material to determine how and why it may resonate with audiences—a process that now falls under the broad banner of the word curation.

人们很容易忽视这样一个事实：当我们通过数字平台消费内容时，我们在特定时刻看到的内容更直接地由方程式而非潮流引领者决定。Netflix 的主页、Facebook 的信息流和 Spotify 的自动广播，并没有来自编辑、DJ 或预订员的直接影响，而是对众包数据的数学处理，这些数据会延伸到涵盖网站上的每一位用户。策展本身是自动化的，规模不断扩大，直到超出任何个人的掌控范围。通过算法信息流，我们获得的只是亚马逊的零售体验，而不是麦克纳利的策展眼光。

It’s easy to overlook the fact that when consuming content through digital platforms, what we see at a given moment is determined more directly by equations than such tastemakers. With Netflix’s home page, Facebook’s feed, and Spotify’s automated radio, there is no direct influence from editor, DJ, or booker, but, rather, a mathematical processing of crowdsourced data stretching to encompass every user on the site. The curation, such as it is, is automated, scaled up until it’s beyond the grasp of any individual person. Through our algorithmic feeds, we get only the Amazon retail experience, not the McNally curatorial eye.

好品味

GOOD TASTE

品味引领者所说的“品味”指的是个人偏好，是我们每个人用来判断自己喜欢什么的洞察力，无论是在音乐、时尚、美食还是文学方面。我们不断地决定听什么、读什么，或者穿什么。这些选择是私密的，反映了我们转瞬即逝的情绪，以及我们个人感知——自我意识——的缓慢形成。

The “taste” of tastemakers means personal preference, the discernment that we all apply to figure out what we like, whether in music, fashion, food, or literature. We make constant decisions to listen to, read, or wear one thing instead of another. These choices are intimate, reflecting our ephemeral moods and the slow building of our individual sensibilities—of our senses of self.

每个人都有自认为“品味高雅”的朋友。比如我的朋友马克，他是一位剧院设计师，但他对音乐的了解令人羡慕——这不是他的工作，而是他的热情所在。虽然很多人在青春期就放弃了音乐的习惯，只是重复听以前喜欢的歌，但马克会听每一个热门的新乐队，看看是否值得与朋友们分享。每隔几个月，我都会向他征求一些建议。我并不总是喜欢他推荐的所有专辑，但我知道这些音乐中总有值得一听的地方。我相信他的判断，而且他非常了解我的个人品味，能够判断哪些音乐最适合我。

Everyone has friends who they think of as having “good taste.” My friend Mark, for instance, works as a theater designer but also has an enviable knowledge of music—it’s not his job; it’s his passion. Though many people leave the habit behind in adolescence and just recycle their past favorites, Mark listens to each buzzy new band and determines whether it’s worth sharing with friends. Every few months, I ask him for a handful of suggestions. I don’t always like all the albums he picks, but I know there’s something worthwhile to listen for in the music. I trust his judgment, and he knows enough about my personal taste to determine which music might suit me best.

在其他情况下，我们会遇到一个朋友，他总是知道带什么酒去吃饭，会关注最热门的时尚品牌，或者会推荐值得一看的电视节目。品味是我们衡量文化和判断我们与文化关系的一个词。如果某样东西符合我们的品味，我们就会对它产生亲近感并产生认同感，并基于它与其他人建立关系，就像顾客们会对服装标签进行交流（要么喜欢某个品牌，要么讨厌某个品牌）。故意表现出来的坏品味可能和好品味一样引人注目，正如作家拉克斯·金在她的书《俗气》中所描述的那样： “俗气就是快乐。”但从本质上讲，品味是一个更深层的哲学概念。它与道德接壤，代表着对世界上美好事物的天生感知。

In other cases, there’s the friend who always knows the right wine to bring to dinner, the friend tuned into the most relevant fashion brands, or the friend who recommends television shows worth watching. Taste is a word for how we measure culture and judge our relationship to it. If something suits our taste, we feel close to it and identify with it, as well as form relationships with other people based on it, the way customers commune over clothing labels (either loving or hating a particular brand). Intentionally bad taste might be just as compelling as good taste, as the author Rax King described in her book Tacky: “Tackiness is joyfulness.” But in its origins, taste is a much deeper philosophical concept. It borders on morality, representing an innate sense of what is good in the world.

18世纪50年代，法国百科全书中的一篇条目探讨了“品味”这一概念，其中引用了哲学家伏尔泰和孟德斯鸠的注释。他们共同为西方的“品味”概念奠定了良好的基础。伏尔泰写道：“要拥有品味，仅仅看到和知道一件作品的美是不够的。一个人必须感受美，并被它感动。仅仅感受到、被模糊地感动也是不够的：必须辨别出不同层次的情感。” 品味超越了肤浅的观察，超越了将某物定义为“酷”。品味要求体验作品的整体，并评估自己对它的真实情感反应，解析其效果。（品味并非被动；它需要付出努力。）孟德斯鸠是一位公共知识分子，同时也是一位男爵和法官。他撰写了《论品味——关于自然和艺术的主题》一文，这篇文学短文在他1755年去世时仍未完成。这是一篇优美而蜿蜒的文章，探讨了什么能愉悦灵魂。孟德斯鸠认为，品味“不过是一种能力，能够敏锐而迅速地发现每件事物应给予人类的愉悦程度。”

In the 1750s, taste was tackled in a French encyclopedia entry with notes from the philosophers Voltaire and Montesquieu, who together offer a good basis for its Western conception. Voltaire wrote, “In order to have taste, it is not enough to see and to know what is beautiful in a given work. One must feel beauty and be moved by it. It is not even enough to feel, to be moved in a vague way: it is essential to discern the different shades of feeling.” Taste goes beyond superficial observation, beyond identifying something as “cool.” Taste requires experiencing the creation in its entirety and evaluating one’s own authentic emotional response to it, parsing its effect. (Taste is not passive; it requires effort.) Montesquieu, who was a baron and a judge in addition to a public intellectual, contributed “An Essay Upon Taste, in Subjects of Nature, and of Art,” a literary sketch that was left unfinished when he died in 1755. It’s a beautiful, meandering piece of writing on what delights the soul. Taste, according to Montesquieu, “is nothing else but an ability of discovering, with delicacy and quickness, the degree of pleasure which every thing ought to give to man.”

“自然的品味并非理论知识，”孟德斯鸠继续说道，“而是对我们甚至未知规则的快速而精妙的应用。”这句话的后半部分让我印象深刻：品味是一种抽象、难以言喻、不稳定的东西。听音乐或读书的人在体验之前无法真正判断自己是否会喜欢某种东西；艺术作品带来的愉悦感从来都无法保证。因此，当我们遇到一件艺术品时，我们会立即用一些心理原则来评判它，并希望发现其中的美，感到肯定，即使我们无法确切描述这种美是什么，或者我们最初是如何确定它的。品味应该是模棱两可的。正如意大利哲学家乔治·阿甘本在其1979年关于品味的专著中所总结的那样：“品味享受美，但无法解释它。”

“Natural taste is not a theoretical knowledge,” Montesquieu continued, “it’s a quick and exquisite application of rules which we do not even know.” The latter part of that statement strikes me: Taste is an abstract, ineffable, unstable thing. A listener to music or reader of a book cannot truly tell if they will enjoy something before they experience it; pleasure in a piece of art is never guaranteed. So when encountering an artwork, we immediately evaluate it by some set of mental principles, and, hopefully, find the beauty in it, feel affirmed, even if we can’t quite describe what that beauty is or how exactly we determined it in the first place. Taste is supposed to be ambiguous. As the Italian philosopher Giorgio Agamben summarized in his 1979 monograph on taste, “Taste enjoys beauty, without being able to explain it.”

品味是自我的根本组成部分；培养或沉溺于品味意味着构建更坚定的自我意识。它成为身份认同的基础。1906年，日本作家冈仓天心撰写了《茶经》，以此来传承日本的品味，并用英文将其传达给他在美国的朋友和赞助人，其中包括艺术收藏家伊莎贝拉·斯图尔特·加德纳。冈仓在一次关于茶室设计的讨论中指出，艺术本身并非旨在千篇一律或迎合大众：“茶室的建造应该迎合某些个人品味，这体现了艺术活力的原则。” 冈仓讲述了17世纪艺术家小堀远州的一段话。远州在与弟子们交谈时，称赞了一位茶道大师收藏的茶道用具，正是因为很少有人欣赏它们：“伟大的利休只敢于欣赏那些他个人喜欢的物品，而我却不自觉地迎合了大多数人的品味。” Enshiu 嘲笑自己的品味过于主流，算不上真正优秀。然而，迎合“大多数人的口味”或许正是算法推送的唯一目标——基于数据的大多数人。

Taste is a fundamental part of the self; developing or indulging it means constructing a firmer sense of self. It becomes the basis for identity. In 1906, the Japanese writer Okakura Kakuzo wrote The Book of Tea, a way of enshrining Japanese taste and communicating it, in English, to his friends and patrons in the United States, a group that included the art collector Isabella Stewart Gardner. Art itself was not meant to be generic or cater to a broad audience, Okakura argued in a discussion of tearoom design: “That the tearoom should be built to suit some individual taste is an enforcement of the principle of vitality in art.” Okakura recounted the story of a statement by the seventeenth-century artist Kobori Enshiu. Speaking to his disciples, Enshiu complimented a fellow tea master’s collection of tea ceremony implements precisely because few others appreciated them: “The great Rikiu dared to love only those objects which personally appealed to him, whereas I unconsciously cater to the taste of the majority.” Enshiu derided his own taste as too mainstream to be truly great. Yet catering to “the taste of the majority” might be the single goal of algorithmic feeds—a majority based on data.

品味并非完全积极或有效。1930年，日本哲学家九鬼周造写了一篇文章，试图定义了一种名为“iki”（意为“厌世”）的日本文化价值观，它相当于一种都市里的厌世情绪，一种在生活方方面面都明显的矛盾心态。（我长期居住在东京的美国作家朋友W. David Marx将其与新英格兰白人新教徒（WASP）的某些方面进行了比较。）爱情、金钱和美貌都可能轻易失去，也可能轻易得到，而得到并不总是比失去更好。缺席和存在一样值得珍惜。“Iki”被理解为一种更高级的品味形式，”Kuki写道。

Taste is not necessarily wholly positive or efficient. In 1930 the Japanese philosopher Kuki Shuzo wrote an essay attempting to define a Japanese cultural value called iki, which amounted to a kind of urbane world-weariness, a pronounced ambivalence in all aspects of life. (W. David Marx, an American writer and friend of mine long living in Tokyo, compared it to aspects of New England WASPiness.) Love, money, and beauty could all be as easily lost as gained, and gaining may not always be better than losing. Absence must be appreciated as much as presence. “Iki is understood as a superior form of taste,” Kuki wrote.

孟德斯鸠至关重要地指出，惊喜——就像一件极其丑陋的侘寂日本茶具——是品味的一个基本要素，它既可能令人感到疏离，也可能令人感到挑战。“某些事物会让我们感到惊讶，是因为它激发了我们的好奇心，或者因为它是新的或意想不到的，”他写道——它存在于我们已知的喜好之外。“当我们的灵魂感受到无法分析的事物，或者当一个物体看起来与它所认知的截然不同时，它常常会体验到愉悦。”理解这种惊喜的感觉需要时间。品味并非瞬间产生，它会随着你对一件艺术品的思考和消化而变化：“当某件事物激发出一种起初只是轻微的惊讶，但这种惊讶会持续、增强，最终转化为赞叹时，我们就会意识到它拥有无比的美。”

Montesquieu crucially argued that surprise, which can be alienating or challenging, like a particularly ugly wabi-sabi Japanese tea vessel, is a fundamental element of taste. “Something can surprise us because it excites wonder, or because it is new or unexpected,” he wrote—it exists outside the realm of what we already know we like. “Our soul often experiences pleasure when it feels something it cannot analyze, or when an object appears quite different from what it knows it to be.” Understanding this feeling of surprise can take time. Taste is not necessarily instantaneous and changes as you consider and digest the experience of an artwork: “We become aware of the presence of great beauty when something inspires us with a surprise which at first is only mild, but which continues, increases, and finally turns into admiration.”

孟德斯鸠引用了文艺复兴时期意大利画家拉斐尔的作品，描述了一件充满力量的艺术作品是如何慢慢燃烧的，其优雅或许会从最初的微妙中意外地迸发出来。对我来说，这让我想起了弗兰克·奥申2016年的专辑《金发女郎》（Blonde），这张专辑发行于他首张专辑《橙色频道》（Channel Orange）发行四年后。起初，我忽略了它。这些曲目听起来不像一首独立的歌曲，更像是一团合成音效，歌词模糊得令人难以捉摸，情感只是被遮蔽和自动调谐。但随着我继续聆听，被音乐中某种难以言喻的特质所吸引，我逐渐意识到，这张专辑的重点在于其抽象性，它的难以捉摸描绘了现代人的疏离感，以及尽管如此，人们仍然需要继续生活下去。当然，《金发女郎》是21世纪初的一部重要的流行杰作，也是一部畅销书。但这张专辑和这位音乐家都没有遵循算法推送的规则。

Citing the work of the Renaissance Italian painter Raphael, Montesquieu described the slow burn of a powerful artwork, whose elegance might emerge unexpectedly from initial subtlety. For me, it brings to mind Frank Ocean’s 2016 album Blonde, which was released four years after his debut album Channel Orange. At first, I overlooked it. The tracks didn’t sound like individual songs, more a wash of synthesized sound, and the lyrics were vague to the point of inscrutability, emotion only veiled and auto-tuned. But as I kept listening, pulled along by some undefinable quality within the music, I came to realize that the album’s abstraction was the point, its elusiveness a portrait of modern alienation and the need to keep living despite it. Of course, Blonde was a major popular masterpiece of the early twenty-first century, a bestseller. But the album and the musician alike didn’t play by the rules of algorithmic feeds.

如果品味确实需要深刻的感受，需要时间去体验，并且受益于陌生事物带来的惊喜，那么技术似乎不可能复制它，因为算法信息流违背了这些基本特质。当推荐算法仅仅基于你和其他平台用户已有的喜好数据时，这些算法就无法提供孟德斯鸠所描述的那种可能不会立即带来愉悦的惊喜。信息流结构还会阻止用户在任何一条内容上花费太多时间。如果你发现某些内容很无聊，甚至太过低调，你就会一直往下看，根本没有时间培养更深层次的欣赏之情——人们越来越倾向于在所有事情上变得急躁和肤浅。正如韩国哲学家韩秉哲在其 2017 年出版的《蜂群》一书中所论述的那样，如此多的人在网上毫无障碍地相互接触——互联网的“去媒体化”——使得“语言和文化变得扁平化和粗俗化”。

If taste indeed must be deeply felt, requires time to engage with, and benefits from the surprise that comes from the unfamiliar, then it seems that technology could not possibly replicate it, because algorithmic feeds run counter to these fundamental qualities. When recommendation algorithms are based only on data about what you and other platform users already like, then these algorithms are less capable of providing the kind of surprise that might not be immediately pleasurable, that Montesquieu described. The feed structure also discourages users from spending too much time with any one piece of content. If you find something boring, perhaps too subtle, you just keep scrolling, and there’s no time for a greater sense of admiration to develop—one is increasingly encouraged to lean into impatience and superficiality in all things. As the Korean philosopher Byung-Chul Han argued in his 2017 book In the Swarm, the sheer exposure of so many people to each other online without barriers—the “demediatization” of the Internet—makes “language and culture flatten out and become vulgar.”

培养自己的品味，也就是一套潜意识中用来识别自己喜好的原则，与被动地消费信息流推送的内容相比，是一项艰巨的挑战。但这种情况不能完全归咎于算法的存在。如今，我们拥有比以往任何时候都更多的文化选择，而且这些选择可以按需获取。我们可以自由地选择任何东西。然而，我们经常做出的选择是没有选择，任由自动化信息流塑造我们的视野，而这些信息流可能基于人类的集体行为，但其本身并非人类的产物。

Building your own sense of taste, that set of subconscious principles by which you identify what you like, is an uphill battle compared to passively consuming whatever content feeds deliver to you. But the situation can’t solely be blamed on the presence of algorithms. Today we have more cultural options available to us than ever and they are accessible on demand. We are free to choose anything. Yet the choice we often make is to not have a choice, to have our purview shaped by automated feeds, which may be based on the aggregate actions of humans but are not human in themselves.

从某种程度上来说，这种向算法的转变很方便。不停地审视自己的偏好很累人：研究有哪些新的文化产品可以买到；阅读杂志或向朋友征求书籍推荐；还要决定吃什么、去哪里吃。这是一种奢侈的劳动，18世纪的法国哲学家或许有充足的时间去做，但在节奏快得多的当代世界，我们大多数人都负担不起。（孟德斯鸠没有Instagram来分散他对拉斐尔画作的注意力。）例如，遵循Netflix主页的建议，就提供了一条捷径。

In some ways, this shift to algorithms is convenient. It’s tiring to interrogate your preferences all the time: researching which new cultural products are available to you; reading magazines or requesting book suggestions from friends; and making decisions about what and where to eat. It’s a luxury form of labor that eighteenth-century French philosophers may have had plenty of time for, but in the much faster-paced contemporary world, most of us cannot afford. (Montesquieu didn’t have Instagram to distract him from contemplating Raphael canvases.) Following the suggestions of the Netflix home page, for example, offers a shortcut.

品味也可能更像是一种担忧，而非个人成就感的来源。基于个人品味做出的选择，如果无意中与他人的喜好相冲突，可能会令人尴尬。我会根据当时的情境选择一些常规的服装，比如穿着运动休闲服去办公室，或者穿着鲜艳的颜色去参加严肃的葬礼。我经常会为一群朋友挑选一家之前从未去过的酒吧或餐厅，我以为这个选择会很受欢迎，但结果却完全不符合我的口味，这让我感到很尴尬。（华盛顿特区有一家这样的酒吧，墙上挂满了动物标本，让人很不舒服。）在这种情况下，Yelp 或谷歌地图的自动推荐可能更适合我：民主、平均的认可减轻了做出可能过于古怪的选择的压力。但与此同时，我也不希望那些最低标准来决定我读什么书或看什么电视节目。文化不是一台可以五星好评的烤面包机——尽管现在被亚马逊收购的网站 Goodreads 试图将这些评分应用于书籍。有很多我喜欢的经历——比如像 Rachel Cusk 的《Outline》这样一部没有情节的小说——其他人无疑会给出差评。但这些是 Filterworld 对所有事物强制执行的规则。

Taste can also feel more like a cause for concern than a source of personal fulfillment. A selection made based on your own personal taste might be embarrassing if it unwittingly clashes with the norms of the situation at hand, like wearing athleisure to the office or bright colors to a somber funeral. I find myself mortified when I pick out a previously unexplored bar or restaurant for a group of friends, a choice I think will be a crowd-pleaser, and it turns out to have entirely the wrong vibe. (One such bar in D.C. turned out to have too many taxidermy animal heads on the walls for comfort.) In that situation, an automated recommendation from Yelp or Google Maps may have suited me better: the proof of democratic, average approval takes the pressure off making a choice that may prove too quirky. Yet at the same time, I wouldn’t want those lowest-common-denominator rules to determine which books I read or television shows I watch. Culture isn’t a toaster that you can rate out of five stars—though the website Goodreads, now owned by Amazon, tries to apply those ratings to books. There are plenty of experiences I love—a plotless novel like Rachel Cusk’s Outline, for example—that others would doubtless give a bad grade. But those are the rules that Filterworld enforces for everything.

品味需要惊喜，它也在挑战和风险中蓬勃发展，在特定方向上走得太远。安全或许可以避免尴尬，但也令人乏味。在二十世纪，品味不再是一个关于艺术品质的哲学概念，而更像是一种与工业时代消费主义平行的东西，一种判断买什么，并根据别人的购买行为来评判他人的方式。这种现象——过于迎合大众口味，从而让自己无法与更具启发性、更个性化的文化接触——在乔治·佩雷克1965年的中篇小说《物》中有所描述。故事讲述了一对二十出头的夫妇，杰罗姆和西尔维，他们是市场调研员，他们采访消费者时会问诸如“为什么纯吸力吸尘器卖得这么差？出身卑微的人对菊苣有什么看法？”之类的问题。这对夫妇是人类数据收集者。他们的个人欲望也类似于市场调查的结果：他们喜欢那些他们应该喜欢的东西。佩雷克为他们虚构的理想公寓绘制的草图里，有玉石烟灰缸、藤椅、茹伊印花布壁纸、瑞典灯具，还有保罗·克利的画作。我承认，现在听起来依然很美好。

As taste requires surprise, it also thrives on challenge and risk, treading too far in a particular direction. Safety may avoid embarrassment, but it’s also boring. Over the twentieth century, taste became less a philosophical concept concerning the quality of art than a parallel to industrial-era consumerism, a way to judge what to buy and judge others for what they buy in turn. This phenomenon—conforming too much with popular taste and thus insulating yourself from having a more inspiring, personal encounter with culture—is depicted in Georges Perec’s 1965 novella Things. The story concerns a couple in their early twenties, Jerome and Sylvie, who work as marketing researchers, interviewing consumers with questions like “Why are pure-suction vacuum cleaners selling so poorly? What do people of modest origin think of chicory?” The couple are human data collectors. Their personal desires also resemble the results of a marketing survey: they like what they are supposed to like. Perec’s sketch of their fictional aspirational apartment includes jade ashtrays, cane-seated chairs, Toile de Jouy wallpaper, Swedish lamps, and Paul Klee drawings. I admit, it still sounds nice to me today.

然而，完美的形象必然包含着某种程度的空虚。当品味过于标准化时，它就会被贬低。“他们摇摆不定的品味，“他们过于犹豫不决、一丝不苟，缺乏经验，对自己认为的真正良好品味的标准过于狭隘，这些给他们带来了一些不和谐的时刻，一些羞辱，”佩雷克写道。比如，杰罗姆试图通过打扮得像个英国绅士来追随当时的时尚，但却只能将其打造成一个“欧洲大陆的漫画”，看起来像个“拿着微薄薪水的新移民”。他还因为穿得太频繁而把一双优雅的英国鞋磨破了。杰罗姆和西尔维知道自己应该喜欢什么，但他们不太明白为什么或如何喜欢。没有品味的消费只是纯粹的、加速的资本主义。

Yet the image of perfection entails a degree of emptiness. When taste is too standardized, it is degraded. “Their still-wavering taste, their over-hesitant meticulousness, their lack of experience, their rather blinkered respect for what they believed to be the standards of true good taste, brought them some jarring moments, some humiliations,” Perec wrote. Such as when Jerome attempts to follow the fashion of the day by dressing like an English gentleman but succeeds only in a “continental caricature” of it, appearing like a “recent emigrant on a modest salary.” He also wears through a pair of elegant British shoes by wearing them too often. Jerome and Sylvie have an idea of what they should like, but they don’t quite understand why or how. Consumption without taste is just undiluted, accelerated capitalism.

两种力量塑造着我们的品味。正如我之前所述，第一种力量是我们独立追求个人喜好，第二种力量是我们对大多数人似乎喜欢什么的认知，也就是主流。两者可能朝着相反的方向发展，但通常更容易追随后者，尤其是在互联网将其他人的消费内容如此迅速地公开的情况下。（如果你没有发布过相关内容，那你真的看过电视节目吗？）算法推送进一步强化了主流的存在，我们的个人选择将以此为依据进行评估。品味无处不在；它涉及“日常生活中最日常的选择，例如烹饪、服装或装饰”，法国社会学家皮埃尔·布迪厄在其1984年出版的《区分：品味判断的社会批判》一书中写道。这些选择可以象征着我们审美偏好之外的一系列事物，例如经济阶层、政治意识形态和社会身份。“品味进行分类，它也将分类者进行分类，”布迪厄写道。难怪我们会担心喜欢什么，有时发现将这个责任交给机器更简单。

There are two forces forming our tastes. As I described previously, the first is our independent pursuit of what we individually enjoy, while the second is our awareness of what it appears that most other people like, the dominant mainstream. The two may move in opposite directions, but it’s often easier to follow the latter, particularly when the Internet makes what other people are consuming so immediately public. (If you didn’t post about it, did you really watch a TV show?) Algorithmic feeds further reinforce the presence of that mainstream, against which our personal choices are evaluated. Taste is inescapable; it involves “the most everyday choices of everyday life, e.g., in cooking, clothing, or decoration,” the French sociologist Pierre Bourdieu wrote in his 1984 book Distinction: A Social Critique of the Judgement of Taste. These choices can be symbolic of a range of things beyond just our aesthetic preferences, such as economic class, political ideology, and social identity. “Taste classifies, and it classifies the classifier,” Bourdieu wrote. No wonder that we worry about what to like, and sometimes find it simpler to export that responsibility to machines.

亚马逊打造了一款旨在迎合大众品味的设备。它被称为 Amazon Echo Look，可以帮你做出所有时尚决定。2017 年 Echo Look 上市后，我就试用了它。Echo Look 是一个白色的塑料小圆柱体，由一根小臂支撑，中央有一个摄像头，就像独眼巨人的独眼一样。它的广告宣传是，它可以轻松拍出你的整套服装——只需将设备放在架子上，然后使用语音指令让它拍照，并摆出轻松的姿势。照片随后会被发送到上传到你的手机，并存储在一个构建了你的衣橱百科全书的应用程序中——你可以像浏览私人 Instagram 账户一样浏览自己的衣服（就像《独领风骚》中的 Cher Horowitz 一样）。但摄像头还会通过一项名为“Style Check”的功能来评判你的衣服，该功能结合了算法分析和人工操作（回想一下亚马逊的仿自动化 Mechanical Turk），告诉你这些衣服是否搭配或者组合是否时尚。

Amazon created a device that was meant to approximate taste. It was called the Amazon Echo Look, and it could make all your fashion decisions for you. I tried it out after it launched in 2017. The Echo Look was a small white cylinder made of plastic raised up on a little arm, with a camera at its center like the lone eye of a cyclops. It was advertised as a means to take easy selfies of your entire outfit—simply stand the device on a shelf, then use voice commands to tell it to take a photo, while jauntily posing. The photo was then sent to your phone and stored in an app that built an encyclopedia of your wardrobe—you could scroll through your own clothes as if they were a private Instagram account (not unlike Cher Horowitz in Clueless). But the camera also judged your clothes with a function called “Style Check,” using a combination of algorithmic analysis and human workers (recall Amazon’s faux-automated Mechanical Turk) to tell you if the garments matched or if the combination was fashionable.

为了了解 Echo 如何评估时尚感，我尝试穿了两套不同的 T 恤和牛仔裤——一套全黑，另一套全灰——站在 Echo Look 前，就像我妈妈在小学拍照日之前那样。我用 Style Check 对比了这两套搭配，黑色衣服的得分更高，时尚度达到了 73%（满分 100 分），而灰色衣服只得到了剩下的 27%。系统几乎没有给出分数的解释，只是说“你搭配这些衣服的方式看起来更好”。也许单色黑色是既定的时尚风格，而单色灰色不那么赏心悦目，因此更激进？我也说不准。

To see how the Echo evaluated style, I tried wearing two different T-shirt and jeans outfits—one set all black and the other all gray—standing in front of the Echo Look like it was my mother before picture day at elementary school. When I compared the two ensembles with Style Check, the black clothes fared better, scoring 73 percent stylish, out of the possible 100, while the gray outfit got only the remaining 27 percent. The system gave little explanation for the score: “The way you styled those pieces looks better,” it said. Maybe monochrome black is an established fashion trope, whereas monochrome gray is less pleasing and thus more radical? I couldn’t tell.

其他精辟的评价包括“更好的颜色搭配”和“衣服的版型更好”。Style Check 还告诉我，穿牛津衬衫时，卷起袖子比扣在手腕处更好，而且竖起衣领是个好主意（这与我自己的品味认知相反，我的品味是在中学时期形成的，当时我反对 Abercrombie 的时尚巅峰）。根据 Echo Look 的算法，蓝色牛仔布是牛仔裤的最佳选择。它提供了评估，但没有提供整体的理解或情感基础。Style Check 只是将您的选择与其档案中的净平均值进行比较。当您的时尚品味与其他人一致时，它就是最好的。此外，Echo Look 还提供符合其理想着装规范的服装的即时购买服务，当然，这些服装由亚马逊销售，亚马逊受益于其对算法平均性的愿景。

Other pithy judgments included “Better color combination” and “The shape of the outfit works better.” Style Check also informed me that when wearing an oxford shirt, rolled-up sleeves are better than buttoned at the wrist, and popping your collar is a good idea (contrary to my own perception of taste, established in middle school rebelling against the peak of Abercrombie fashion). Blue denim is the best choice for jeans, according to Echo Look’s algorithm. It offered evaluation, but no holistic understanding nor grounding in emotion. Style Check simply compared your choices to the net average of data in its archives. Your taste in fashion was best when it followed everyone else’s. What’s more, the Echo Look also offered instant purchases of clothing that matched its ideal dress code, sold by Amazon, of course, which profited from its vision of algorithmic averageness.

这是一种自下而上的文化偏好模式，既与个人品味的定义相悖，也与互联网时代之前的“品味引领者”体系相悖。“品味引领者”们挑选出最酷的东西，并将其强加于人。这种等级制度可以用一个2006年电影《穿普拉达的女王》中的一幕——如今也成了网络红人。梅丽尔·斯特里普饰演的《Vogue》主编安娜·温图尔简直就是翻版，而安妮·海瑟薇则是她天真的助理，刚刚开始学习时尚媒体的门道。片中，海瑟薇穿着一件厚重的蓝色毛衣，这件毛衣是她在百货商店打折时一时兴起买的，看起来与其说是因为它时尚，不如说是因为它方便。但斯特里普告诉她，这个决定是像她一样的时尚编辑们早就替她决定的。斯特里普的独白傲慢无礼：“那蓝色代表着数百万美元和无数的工作机会，你以为自己做出了一个让你免于时尚界的选择，而实际上，你穿的这件毛衣是这房间里的人从一堆‘东西’里帮你挑选出来的，这真是有点滑稽。”斯特里普说。这件毛衣是人类的时尚引领者挑选出来的。

This is a bottom-up model of cultural preferences at odds with both the personal definition of taste and the pre-Internet system of tastemakers, individuals who handpicked what was cool and imposed it on everyone else. The hierarchy is best depicted in a scene—and now meme—from the 2006 film The Devil Wears Prada. Meryl Streep plays a facsimile of the Vogue editor in chief Anna Wintour, while Anne Hathaway is her naive assistant who is just learning the ropes of fashion media. In one moment, Hathaway wears a chunky blue sweater that she picked up on a whim at a department store sale, seemingly less for its stylishness than for its convenience. But Streep tells her that the decision was dictated for her well in advance, by fashion editors like herself. Streep’s monologue is imperious: “That blue represents millions of dollars and countless jobs, and it’s sort of comical how you think you made a choice that exempts you from the fashion industry when, in fact, you’re wearing a sweater that was selected for you by the people in this room from a pile of ‘stuff,’ ” Streep says. Human tastemakers picked it out.

归根结底，品味本身就是一种算法，它能将事物分为好坏。它基于个人偏好、从市场营销中吸收的先入之见、社会象征意义以及对某种文化的直接体验等因素，最终会得出一个关于你对眼前事物是感到愉悦还是厌恶的个人答案。因此，区分这种有机的社会代码与推荐算法的软件代码可能很困难，但区分这一点至关重要。

Taste, in the end, is its own kind of algorithm that classifies things as good or bad. The equation that it is based on factors in personal preferences, preconceptions absorbed from marketing, and social symbolism, as well as the immediate experience of a piece of culture, and eventually produces a personal answer for whether you find the thing at hand enjoyable or repulsive. So it can be difficult to distinguish that organic social code from the software code of recommendation algorithms, though it is vital to do so.

应该由人类时尚编辑告诉你喜欢什么，还是应该由算法机器来告诉你，比如亚马逊书店、Spotify 信息流或 Netflix 主页？这就是 Filterworld 文化的核心困境。前者反复无常，由精英守门人推动，他们是一个强大的群体，在一个世纪的现代文化产业中积累起来，充斥着自身的盲点和偏见，包括性别和种族偏见。（这个群体不仅包括纽约市以白人为主的时尚杂志编辑，还包括好莱坞制片人、唱片公司高管和博物馆策展人。）然而，在算法生态系统中，当大众的行为决定了哪些内容容易被看到时，人类的缺陷可能会变得更加明显。种族主义、性别歧视和其他形式的偏见实际上是这个等式的一部分。

Should the human fashion editor tell you what to like or should it be the algorithmic machine, in the form of the Amazon bookstore, Spotify feed, or Netflix home page? That is the central dilemma of culture in Filterworld. The former option is mercurial and driven by elite gatekeepers, a powerful group built up over a century of modern cultural industries, riddled with their own blind spots and biases including those of gender and race. (That group includes not just the mostly white fashion magazine editors of New York City but also Hollywood producers, record-label executives, and museum curators.) Yet the human flaws may become even more dramatic in an algorithmic ecosystem when the actions of mass audiences dictate what can easily be seen. Racism, sexism, and other forms of bias are a de facto part of that equation.

在网上，用户往往被孤立，无法接触到与自身观点和文化相冲突的事物。整个数字环境被科技公司所主导，这些公司有着残酷的资本主义扩张主义动机，这些并不能为文化提供最肥沃的土壤。杂志时尚编辑或许会定期运用他们的能力来挑选和推广一些从未被听到的声音，但算法推送永远都不会这样做；它只能在既定的互动基础上进行迭代。我们用户很少有机会遇到令人震惊的新事物并自己决定是否喜欢它。举例来说，时尚作为一种艺术形式，往往最强大的地方在于它不墨守成规、不追逐平均水平。它的吸引力之一在于打破社会规范：穿着出人意料或奇特的衣服，有时甚至挑战自己的品味。这是任何自动推荐都无法企及的。算法推送是一把双刃剑：边缘化的时装设计师可能会找到一种方法来钻Instagram算法的空子，从而提升自己的人气，而无需等待可能对他们有偏见的白人编辑的注意。但他们这样做也是为了迎合一家比编辑更强大、更狭隘的科技公司的信条。

Online, users are often insulated from views and cultures that clash with their own. The overall digital environment is dictated by tech companies with ruthlessly capitalist, expansionary motives, which do not provide the most fertile ground for culture. While the magazine fashion editor may periodically use their ability to pick out and promote a previously unheard voice, the algorithmic feed never will; it can only iterate on established engagement. We users have less chance of encountering a shockingly new thing and deciding for ourselves if we like it. Fashion, to take one example, is often strongest as an art form when it doesn’t follow the rules and chase averages. Part of its appeal lies in breaking with the social code: wearing something unexpected or strange, even at times challenging your own taste. It’s something that no automated recommendation alone can approximate. Algorithmic feeds are a double-edged sword: A marginalized fashion designer might find a way to game the Instagram algorithm and spark their own popularity without waiting to be noticed by a white editor who might be biased against them. But they are then conforming to the tenets of a tech company even more powerful and more blinkered than the editor.

亚马逊 Echo Look 或许在算法层面上走得太远了。评测者们欣赏这款设备的创新，认为它更像是一台普通的相机，而非一款造型辅助工具。但它最终未能成为一款热门产品。2020 年，亚马逊宣布将停产这款设备，相机及其配套应用也将完全停止运营。一份相关声明称，该公司的目标是“将人工智能和机器学习应用于时尚领域”。Style Check 功能已集成到亚马逊购物应用中。或许，它最终的目标是收集我们的品味数据，以便在未来的应用中结合完善的推荐算法加以利用，而这一目标最终得以实现。

The Amazon Echo Look may have been an algorithmic bridge too far. Reviewers appreciated the device’s innovation, more as a plain camera than a style aid. But it never became a popular product. In 2020, Amazon announced that the device was going to be discontinued and the camera and attendant app would cease functioning entirely. The company aimed to “apply AI and machine learning to fashion,” an attendant statement said. The Style Check function was integrated into the Amazon shopping app. Perhaps its ultimate goal of collecting data about our tastes, to be leveraged in some future application with a perfected recommendation algorithm, was achieved after all.

味觉的正常化

THE NORMALIZATION OF TASTE

算法压力的力量并非理论上的。它并非阴郁的反乌托邦未来，而是一种无处不在的力量，已经影响着文化消费者和创造者。在消费者方面，大量的推荐轰炸会引发一种催眠状态，使收听、观看或购买产品几乎成为必然——无论它是否真正符合你的品味。我注意到，在一篇由……发表的建议专栏中，就发生了这种情况。时尚评论家 Rachel Tashjian 在她的电子邮件通讯《奢华秘诀》中写道。二十出头的 Valerie Peter 曾在 2022 年写信给 Tashjian，抱怨算法推送的信息让她更难找到自己的风格偏好。“我上网已经十年了，我不知道自己喜欢什么，还是算法希望我喜欢什么，”Peter 写道，表达了她对算法的严重焦虑。Instagram、TikTok 和 Pinterest 都感觉像是死胡同。“我想要真正喜欢的东西，而不是低调推销给我的东西，”她在信的结尾写道。我联系了 Peter，想了解这场品味危机的真正原因，最终我们讨论了社交媒体的兴起如何从根本上改变了我们与文化的关系。

The force of algorithmic pressure is not theoretical. It’s not a gloomy dystopian future but, rather, a pervading force that is already influencing cultural consumers and creators. On the consumer side, the bombardment of recommendations can induce a kind of hypnosis that makes listening to, watching, or buying a product all but inevitable—whether it truly aligns with your taste or not. I noticed precisely that happening in an advice column published by the fashion critic Rachel Tashjian in her email newsletter Opulent Tips. Valerie Peter, a woman in her early twenties, had written to Tashjian in 2022, complaining that algorithmic feeds had made it harder to figure out her own style preferences. “I’ve been on the internet for the last 10 years and I don’t know if I like what I like or what an algorithm wants me to like,” Peter wrote, expressing an acute case of algorithmic anxiety. Instagram, TikTok, and Pinterest all felt like dead ends. “I want things I truly like, not what is being lowkey marketed to me,” she ended her letter. I got in touch with Peter to find out what exactly had caused this crisis of taste, and we ended up discussing how the rise of social media fundamentally changed our relationship to culture.

当时，彼得正在读电气工程研究生，住在她上学的地方——英国曼彻斯特。小时候，她辗转于尼日利亚和英国之间，对时尚有着浓厚的兴趣，关注着所有的时装秀。朋友们告诉她，她应该成为一名时尚作家，但她决定走一条经济更稳定的路线，把这份热情留给个人享受。她从小就开始上网，并在2011年青少年时期加入了Facebook。渐渐地，社交媒体让她感到无法逃避；尤其是在疫情期间，彼得开始严重依赖它来与外界联系。“它也渐渐渗透到了现实生活中，”彼得说。最近，潮流周期加速演变成“微趋势”，几周内就会出现又消失；当朋友们引用她没看过的表情包或视频时，她会觉得自己错过了什么。（无法跟上算法的焦虑。）

At the time, Peter was finishing graduate school in electrical engineering and was living where she went to school, in Manchester, England. As a child, moving between Nigeria and Britain, she had been deeply interested in fashion and followed all the runway shows. Friends told her she should be a fashion writer, but she decided to pursue a more financially stable route and keep her passions for personal enjoyment. She had been on the Internet since an early age and joined Facebook in 2011, as a tween. Gradually, social media had come to feel inescapable; particularly during the pandemic, Peter began relying heavily on it as a connection to the outside world. “It has kind of slipped into real life as well,” Peter said. Lately, trend cycles have accelerated into “microtrends” that come and go in a matter of weeks; she feels like she’s missing out when friends cite a meme or video she hasn’t seen. (The anxiety of not keeping up with the algorithm.)

2021年末，彼得被一股微潮流所吸引。一种毛茸茸的、拉到膝盖的暖腿套，突然占据了她的动态——她的Instagram探索页面、TikTok“为你推荐”动态和Pinterest推荐，一下子就占据了她的所有动态。这不是付费广告，却无处不在。在她在网上接触到这些暖腿套之前，彼得告诉我：“我从来不会想到暖腿套。我觉得它们丑陋、丑陋、荒谬。” 但很快，她发现自己几乎是下意识地一时兴起，就在网上点击了一下按钮就买了一双——“神奇地”，她说。这次购买最终并没有改变她的想法；彼得只穿了这双倒霉的暖腿套一小会儿。她把它们藏在衣柜深处，之前也穿过几次。她说，买这些衣服“我甚至不确定自己是否做过这个选择”。

In late 2021, Peter got caught up in one microtrend. A fad of leg warmers, fuzzy fabric tubes pulled up to the knee, suddenly took over her feeds—her Instagram explore page, TikTok “For You” feed, and Pinterest recommendations, all at once. It wasn’t paid advertising, and yet it was everywhere. Before she was exposed to them online, Peter told me, “I would never think about leg warmers. I thought they were ugly, hideous, ridiculous.” But soon enough she found herself buying a pair online with the click of a button, on an almost subconscious whim—“magically,” she said. The purchase ultimately didn’t change her mind; Peter wore the ill-fated leg warmers only a few times before she stashed them at the back of her closet. Buying them was “a choice that I’m not even sure I made,” she said.

她认为，算法的影响在社交媒体信息流中具有侵入性，而她认为这些信息流大多是自然生成的，就像赤裸裸的产品植入会干扰一部原本引人入胜的电影一样。然而，就像我在世界各地遇到的那些“普通咖啡店”一样，没有任何一个实体强迫所有网红和平台都顺应潮流；拥抱暖腿套只是一种利用信息流、获得更多推广、从而获得更多关注和粉丝的方式。只要彼得参与了哪怕只是一条关于暖腿套的帖子，推荐就会蜂拥而至，内容就变得无处可逃。（这遵循着“如果你喜欢这件东西，你肯定会越来越喜欢它”的算法逻辑。）

The algorithmic influence had felt invasive in social media feeds that she believed were largely organic, the way blatant product placement might intrude on an otherwise absorbing movie. Still, just like the Generic Coffee Shops I encountered around the world, no one entity had forced all the various influencers and platforms to conform to the trend; embracing leg warmers was simply a way to game the feed, to get more promotion and thus attention and followers. Once Peter engaged with so much as a single post about the leg warmers, the recommendations pounced and the content became inescapable. (Following the algorithmic logic of “if you like this one thing, you’ll definitely like more and more of it.”)

彼得在梵克雅宝 (Van Cleef & Arpels) 珠宝和占星术方面也有过类似的经历。梵克雅宝珠宝在TikTok上亮相真人秀后迅速走红。彼得在推特上也曾短暂关注过占星术，但后来就失去了兴趣。但推特的推荐系统不断向她推送有关占星术的推文，而且往往是负面警告，无论她多少次要求停止。“每次水星逆行，我都会开始担心自己的生命安全，”她说。“我不想看到它，但我总是被展示出来。这正在毁了我的生活。”这些推送不仅会试图猜测你的喜好，甚至可能无法理解你的偏好何时会发生变化或演变。

Peter had a similar experience with Van Cleef & Arpels jewelry, which became popular on TikTok after it appeared on a reality show, and astrology, a subject Peter briefly followed on Twitter and then lost interest in. But Twitter’s recommendations kept serving tweets about astrology to her, often negative warnings, no matter how many times she requested them to stop. “I started fearing for my life every time Mercury was in retrograde,” she said. “I don’t want to see it, but I keep being shown it. This is ruining my life.” Not only does the feed try to guess what you like, it may not understand when your preferences move on or evolve.

算法焦虑在一定程度上是由定向在线广告的祸害所引发的，这些广告使用的算法与信息流相同。你的参与度会被数字监控追踪，然后你会收到与你互动内容相匹配的产品广告，这些产品来自那些为你的注意力付费的品牌。由于广告是许多数字平台和在线出版物的主要盈利方式，它们无处不在，会打断文章，还会随着自动播放的视频弹出。与电视广告或纸质杂志广告不同，它们是个性化的，会反复出现你可能宁愿忽略的主题——因为许多出版商都通过同样的软件（例如谷歌的AdSense）销售广告，所以同样的广告可能会出现在每个网站上。我曾经在一件家具上体验过这种萦绕心头的体验，那是一个由德国制造商USM制造的餐具柜。这件家具由色彩鲜艳的金属搁板组成的现代主义网格看起来到处都是，不知不觉就吸引了我的眼球。然而，我非但没有更渴望这个餐具柜，反而对它感到厌倦，甚至开始怀疑自己的喜好。

Algorithmic anxiety is fueled in part by the scourge of targeted online advertising, which uses the same kind of algorithms as the feeds. Your engagement is tracked by digital surveillance, and then you are served ads for products that match what you engage with, from brands that pay for your attention. Since ads are the primary way that many digital platforms and online publications make money, they are everywhere, interrupting articles, popping up with autoplay videos. Unlike a television commercial or a print magazine ad, they are personalized, and reiterate subjects you might prefer to ignore—because so many publishers sell their ads through the same software, like Google’s AdSense, the same ads might appear on every website. I’ve experienced this kind of haunting with a piece of furniture, a credenza made by the German manufacturer USM. The piece’s modernist grid of brightly colored metal shelves appears everywhere, catching my eye before I even realize it. Rather than making me desire the credenza more, however, I’ve grown tired of it, suspicious of my own preferences.

“如今，很多文化都是围绕社交媒体塑造的；微潮流层出不穷。你还没来得及眨眼，还没来得及决定是否喜欢某个东西，信息流就已经转移到下一个了，”彼得向我解释道。她总结了这个问题：“我只想知道我喜欢的，是不是我真心喜欢的。”换句话说，她在寻求对自身品味的自信和稳定。虽然她知道自己对推荐负有部分责任，但网络内容的算法加速发展已经远远超越了她的日常体验，达到了荒谬的程度。“并非我在现实生活中的每一次互动都会影响我的选择。为什么我和一个穿着护腿的网红最细微的互动就要影响我的选择？他们把这些都反馈给了我，”她说。在曼彻斯特的人行道上亲身漫步，比起她那些混杂的数字信息流，能感受到更多的多样性和原创性。

“A lot of culture is shaped around social media now; there are so many microtrends. Before you blink or decide if it’s something you like or not, the feed has moved on to the next thing,” Peter explained to me. She summarized the problem: “I just want to know that what I like is what I actually like.” In other words, she was seeking confidence and stability in her own sense of taste. Though she knew she was partially responsible for her recommendations, the algorithmic acceleration of content online outpaced her everyday experiences to an absurd degree. “Not every interaction I have in real life would shape my choices. Why should my tiniest interaction with an influencer wearing leg warmers? They fed it to me,” she said. There was more diversity and originality in a physical walk down the Manchester sidewalk than in her digital feeds, which all ran together.

我的个人设计审美也和我的一样，偏向中世纪和极简主义——毕竟，这些风格都是Instagram的通病。早在接触社交媒体之前，我就对这种风格很感兴趣，但最近，我关注的精选账号从涓涓细流变成了洪流。我的动态里充斥着越来越多不受欢迎的账号推荐，他们发布优雅的室内装饰：墨西哥、瑞典、日本的房屋，干净整洁，米色墙面，绿植点缀。就像彼得的暖腿套问题一样，我喜欢偶尔浏览这类内容，并不意味着我想一直看下去。这些推荐非但没有达到吸引我注意力的目的，反而迫使我直面这种审美缺乏背景和意义。算法推送可能会让我相信我并不喜欢我自以为喜欢的东西，或者至少会因为内容过于饱和而让我更快地改变我的口味，就像一顿饭吃得太频繁，反而会让人觉得没胃口一样。

I feel the same way about my personal design aesthetic, which tends toward the mid-century and minimalist—which are, after all, the generic styles of Instagram. I was already interested in the style before my time on social media, but the slow trickle of curated accounts that I followed has lately turned into a flood. My feed is full of unwanted recommendations of ever more accounts posting elegant room interiors: blandly clean, beige-walled, plant-bedecked homes in Mexico, Sweden, Japan. As with Peter’s leg warmers problem, just because I enjoyed a sprinkling of this content doesn’t mean I want to see it all the time. Instead of accomplishing the goal of sustaining my attention, the recommendations force me to confront the aesthetics’ lack of context and meaning. It may be that the algorithmic feed will convince me that I don’t like what I thought I liked, or at least turn my taste against it much faster than otherwise due to its oversaturation, the way a meal seems less appetizing when you eat it too often.

对彼得作为消费者而言，算法式的品味既无聊又疏离。相比之下，从创作者的角度来看，无处不在的商品反而能带来利润。对于梵克雅宝这样的商业品牌来说，特定产品获得更多关注几乎总是一件好事。产品触达的受众越广，销量就越好。模因越大越好。然而，算法推荐也会扭曲文化创作者发布内容的意图，从而改变他们与自身作品的关系。很多情况下，双方对推广的内容都不满意。

Algorithmic taste, in Peter’s case as a consumer, was both boring and alienating. On the creator side, by contrast, ubiquity can be profitable. For a commercial brand like Van Cleef & Arpels, more attention on a particular product is an almost invariably positive thing. The wider an audience it reaches, the more of a product sells. The bigger the meme, the better. Yet algorithmic recommendations also have a way of warping cultural creators’ intentions for what they put out into the feed, changing their relationship to their own work as well. In many cases, neither side is happy with what is getting promoted.

2018 年初，54 岁的音乐家 Damon Krukowski 是 20 世纪 80 年代独立乐队 Galaxie 500 的鼓手。他注意到 Spotify 上他以前乐队的音乐出现了异常。自从乐队买回音乐版权并将其转移到自己的厂牌后，Krukowski 就能清楚地看到乐队各首曲目的播放量。其中一首名为“Strange”的曲目，出自 1989 年的专辑《On Fire》，播放量远超乐队其他任何曲目——播放量足足高出数十万次。（这也意味着这首歌为 Krukowski 和他的前乐队成员赚了更多钱。）在流媒体播放量排行榜上，“Strange”的播放量与其他曲目呈四十五度角上升。而且这首歌在最初发行时还不是单曲，也没有受到任何营销推广或优先关注。 “对我来说，这完全是偶然的，”克鲁科夫斯基告诉我。更奇怪的是，《Strange》的热潮只在 Spotify 上出现；它在其他流媒体服务上远没有这么火爆。

In early 2018, a fifty-four-year-old musician named Damon Krukowski, the drummer of the 1980s indie band Galaxie 500, noticed something strange was going on with his old band’s music on Spotify. Since the band bought back the rights to its music and moved it to its own label, Krukowski could see exactly how many streams its various tracks were getting. One track, “Strange” from its 1989 album On Fire, was getting played far more often than any other one from the band’s discography—hundreds of thousands more plays. (Which also meant that the song was making more money for Krukowski and his former bandmates.) On a chart of streaming volume, the line for “Strange” diverged from the rest of the catalog, rising at a forty-five-degree angle. And that track hadn’t been a single, had not received any marketing push or prioritization when it originally came out. “To me, it was totally random,” Krukowski told me. Adding to the strangeness was the fact that the “Strange” boom was on Spotify alone; it wasn’t nearly so popular on other streaming services.

对克鲁科夫斯基来说，《奇异》有点像个玩笑：对更流行音乐的粗略模仿。乐队创作这首歌时，他给它起了个绰号叫“重金属民谣”。这首歌在整张低保真、略显笨拙的专辑里显得格格不入。Galaxie 500 的音乐以内敛为傲。他们属于一个不愿抛头露面、充满知识分子气息的乐队，既朋克又不失书呆子气——这三人，加上贝斯手 Naomi Yang 和吉他手 Dean Wareham，在哈佛大学读书时就开始一起演奏了。

For Krukowski, “Strange” was meant as kind of a joke: a loose parody of more popular music. His nickname for the track while the band was composing it was “Heavy Metal Ballad.” It’s an outlier on the otherwise lo-fi, shambling album. Galaxie 500’s music is proudly introverted. They belonged to a scene of spotlight-shy, intellectual bands that were punk but also nerdy—the trio, with bassist Naomi Yang and guitarist Dean Wareham, began playing together as students at Harvard.

Galaxie 500乐队在1991年仅发行了三张专辑后就解散了，但克鲁科夫斯基继续他的音乐和作家生涯，并最终与他的伴侣杨定居马萨诸塞州剑桥。在2010年代管理乐队曲目的商业版权时，他成为了流媒体行业的批评者。唯一可能导致“奇怪”现象的改变是，Spotify在2017年将其自动播放选项设为默认设置。因此，每当用户选择的音乐停止播放时——无论一首曲目、一张专辑或一个播放列表——另一首算法推荐的歌曲会立即播放。

Galaxie 500 broke up in 1991 after releasing just three albums, but Krukowski continued his career as a musician and author and eventually settled in Cambridge, Massachusetts, with Yang, his partner. While managing the commercial rights of his band’s catalog in the 2010s, he became a critic of the streaming industry. The only thing that had changed to possibly cause the “Strange” phenomenon was that in 2017, Spotify had made its autoplay option the default. So anytime the music that the user had chosen stopped—whether a track, album, or playlist—another algorithmically recommended song would instantly play.

“他们切换预设的那一天，就是将一首歌曲从我们目录中分离出来的开始，”克鲁科夫斯基说：“《Strange》被系统推荐的次数比 Galaxie 500 的任何其他歌曲都多。”克鲁科夫斯基在一封时事通讯中发表了他的看法，这引起了当时担任 Spotify“数据炼金术士”的格伦·麦克唐纳的注意。麦克唐纳进行了内部调查，得出结论，《Strange》之所以中了算法的头彩，并不是因为 Galaxie 500 独特的音乐风格，而是因为这首歌与其他乐队的歌曲比与其他曲目更相似。在很多情况下，如果播放《Strange》，听众不太可能点击“跳过”或“停止”按钮，因此推荐系统将其视为有效选择，并提供给更多的听众。

“On the day that they switched the preset, it was the beginning of this separation of one song from the rest of our catalog,” Krukowski said: “Strange” was being recommended by that system more often than any of Galaxie 500’s other songs. Krukowski posted his observations in a newsletter, which attracted the attention of Glenn McDonald, who at the time worked as Spotify’s “data alchemist.” McDonald did an internal investigation and concluded that “Strange” had hit the algorithmic jackpot not because of Galaxie 500’s unique musical style, but because the song was more similar to songs by other bands than Galaxie 500’s other tracks. In many cases, if “Strange” played, the listener was unlikely to hit the Skip or Stop button, and so the recommender system registered it as an effective selection and offered it to ever more listeners.

当我问克鲁科夫斯基为什么认为这首歌如此成功时，他指出这首歌中带有一些幽默的特质，与八十年代流行的重金属民谣颇为相似：鼓点规律的节拍，尖锐的吉他音色（这在《Galaxie 500》中并不常见），缺乏长篇独奏，时长较短，只有三分十九秒。换句话说，《Strange》听起来就像一首普通的歌曲。“一起演奏很有趣，因为我们允许自己比平时多一些批判性，”克鲁科夫斯基说。“这正是算法在歌曲中听到的。”此外，歌曲中还存在一个算法反馈循环，流行的东西会变得更流行。“一旦它被大量推荐，就会得到更多推荐，”他说。考虑到当时 Spotify 推荐系统的优先级，最普通的内容往往获得最佳效果。

When I asked Krukowski why he thought the song hit the mark, he identified the jokey qualities that made it not unlike the heavy-metal ballads that were popular in the eighties: a regular backbeat on the drums, a screechy guitar tone that was uncharacteristic for Galaxie 500, a lack of long solos, and a short run time, at three minutes and nineteen seconds. In other words, “Strange” sounds like a regular song. “It was fun to play together because we were allowing ourselves to bash a little more than usual,” Krukowski said. “That’s what the algorithm heard in the song.” There was also an algorithmic feedback loop, in which what is popular becomes more popular. “Once it’s out there being recommended a whole lot, then it gets recommended even more,” he said. Given the priorities of Spotify’s recommender system at the time, the content that was the most generic succeeded best.

这就是算法标准化的原理。“正常”指的是不引人注目、平庸，只要不会引起负面反应就行。任何符合这个平均水平的内容都会得到加速推广和增长，就像《奇异博士》那样，而其余的内容则会被淘汰。随着越来越少的人看到那些没有得到推广的内容，人们对它们的认知度也会降低，创作者创作这些内容的动力也会随之减弱——部分原因是这些内容在经济上不可持续。（Filterworld 的文化规则是：要么爆红，要么死。）审美可接受性的界限越来越窄，直到只剩下中间的一列。虽然流行风格像移动的目标一样不断变化，但中心化和规范化却始终存在。这就是为什么当某种推文模式开始大获成功时，似乎推特上的每个人都会突然效仿：比如，发布一个开放式的推荐请求，或者把一个荒诞的笑话说成是你孩子（真实的或虚构的）对你说的话。语言本身也被规范化了。“有一种要求正常的压力。这种压力就是：保持原样，任何你熟悉的东西都是安全的，并且不知何故让你觉得自己是群体的一部分，”克鲁科夫斯基说。“这有一个可怕的消失点，那就是法西斯主义。”他将过滤世界的状况描述为“正常的黑洞”。

This is how algorithmic normalization happens. Normal is a word for the unobtrusive and average, whatever won’t provoke negative reactions. Whichever content fits in that zone of averageness sees accelerated promotion and growth, like “Strange” did, while the rest falls by the wayside. As fewer people see the content that doesn’t get promoted, there is less awareness of it and less incentive for creators to produce it—in part because it wouldn’t be financially sustainable. (The rule of culture in Filterworld is: Go viral or die.) The bounds of aesthetic acceptability become tighter and tighter until all that’s left is a column in the middle. While popular styles shift, like moving targets, the centralization and normalization persist. That’s why when a particular pattern of tweet starts succeeding wildly, seemingly everyone on Twitter suddenly copies it: posting an open-ended request for recommendations, for example, or presenting an absurd joke as something your child (real or made up) said to you. Language itself gets normalized. “There’s a pressure to be normal. That pressure is just saying: Be the same, whatever’s familiar to you is safe and somehow makes you feel like part of the group,” Krukowski said. “There’s a horrible vanishing point to that, and that is fascism.” He described the condition of Filterworld as a “black hole of normalcy.”

法西斯主义意味着被迫遵守单一意识形态的世界观，这种世界观可能完全忽视特定身份或人群。这是同质化的强制要求。过滤世界也可能是法西斯主义的，因为算法推送的内容往往会创建事物应有的样子模板，这些模板总是带有固有的偏见——这是一种对现实的括号，而用户创造符合这种模型的内容则满足了这种括号。这种括号不仅包括身份，也包括文化形式。用独裁或封建来形容过滤世界或许也准确：我们都生活在我们无权掌控的网络空间里，遵循着我们并不认可的反复无常的规则。

Fascism means being forced to conform to the tenets of a single ideological view of the world, one that may utterly discount a particular identity or demographic. It is the mandate of homogeneity. Filterworld can be fascistic, in that the algorithmic feeds tend to create templates of how things are supposed to be, always informed by inherent biases—a bracketing of reality that is then fulfilled by users creating content that fits the mold. That bracketing includes forms of culture as well as identities. It might also be accurate to describe Filterworld as dictatorial or feudal: we all reside online within spaces we have no power over, following capricious rules that we don’t approve.

克鲁科夫斯基并不讨厌他迟来的热门歌曲《Strange》，但他说：“我从来没觉得它是一首出彩的歌。如果你眯起眼睛——算法就是这样——把头转向一边，隔着纱布听，你可能会把它误认为是另一支乐队。”这首歌并没有体现出鼓手对Galaxie 500乐队的创作意图的理解；这首歌最初只是一个音乐玩笑，却不知不觉地成为了乐队整体作品的象征。乐队实际上选择了节奏更慢、更安静的《Blue Thunder》作为最初的主打单曲——但一些唱片公司高管并不认同这一选择。

Krukowski doesn’t dislike his belated hit “Strange,” but, he said, “I never thought it was a standout track. If you squint—which is what the algorithm does—and listen with your head turned to one side, through a gauze, you could mistake it for a different band.” The song didn’t represent the drummer’s sense of what Galaxie 500 was trying to achieve creatively; what started as something of a musical joke unwittingly became emblematic of the band’s work as a whole. The band had actually picked the slower, quieter “Blue Thunder” as the original lead single—a choice some of their label executives disagreed with.

在现代算法推荐的推动下，艺术家对于流行音乐的选择越来越少，对自己作品出现的环境的控制也越来越少。Spotify 的界面，该算法会在页面顶部突出显示乐队播放次数最多的歌曲，这使得按原始顺序查找和播放整张专辑变得更加困难。Galaxie 500 无法像反抗唱片公司那样有意抵制算法的决策。Krukowski 继续说道：“算法的奇怪之处在于，无论我们是否愿意，我们都承受着这种压力。即使我们不想与它有任何关系，算法也会突然出现，挑选你最‘正常’的歌曲，并让你认同它，而不是你最特别的‘你’的曲目。”同样的过程也适用于艺术家、作家以及任何在网上感到压力要将自己局限于任意“个人品牌”的人。

With modern-day algorithmic recommendations, artists have much less choice in what becomes popular and even less control over the context that their work appears in. Spotify’s interface, which highlights a band’s most played songs at the top of the page, makes it harder to find and play a full album in its original order. Galaxie 500 can’t intentionally resist the algorithmic decisions like they could push back against their label. Krukowski continued, “The weird thing with algorithms is that we’re all under that pressure, whether we want to be or not. Even if we don’t want anything to do with it, the algorithm is going to swoop in and pick your most ‘normal’ song and make you identified with that, instead of your most peculiarly ‘you’ track.” The same process applies to artists, writers, and anyone online who has felt pressure to confine themselves to an arbitrary “personal brand.”

而且由于 Spotify 严格控制听众与音乐的互动方式，他们不必像大厂牌那样通过丰厚的唱片合约或其他福利来激励音乐人。“为了承受算法的这种压力，他们没有给我们任何回报，”克鲁科夫斯基说。甚至连稳定的生活都没有。《Strange》在 Spotify 上线以来的播放量为 1400 万次，而 Galaxie 500 的第二大歌曲《Tug Boat》的播放量为 900 万次。然而，《Strange》在 Spotify 上线以来的收入仅为 1.5 万美元。用曝光度换取收入，尤其是以牺牲乐队的身份认同为代价，对音乐人来说并不公平：“这是一种榨取。”然而，许多创作者为了融入 Filterworld 不得不做出这样的交易。

And because Spotify implacably controls how listeners interact with the music, they do not have to incentivize musicians in the same way as major labels did, with rich record deals or other perks. “To have this pressure from an algorithm, they’re not offering us anything in exchange,” Krukowski said. Not even a stable living. “Strange” has fourteen million streams in its Spotify lifetime, while Galaxie 500’s second-biggest song, “Tug Boat,” has nine million. Yet the lifetime Spotify earnings of “Strange” have amounted only to fifteen thousand dollars. The trade-off of exposure for revenue, especially at the expense of the band’s sense of identity, doesn’t feel fair to the musician: “It’s extractive.” Yet such is the deal that so many creators have to make to fit in to Filterworld.

Netflix 分类

CATEGORIZED BY NETFLIX

在我成年后的大部分时间里，我都没有电视。大学时，我通常会盗版自己想看的节目，比如《广告狂人》，这种不正当的习惯一直延续到2010年代的《权力的游戏》时代。我大部分的电视节目和电影都是通过经典的观看方式观看的：把笔记本电脑放在床边，头靠着枕头微微向上，以获得完美的观看角度。Netflix 于 2007 年推出了其流媒体服务的早期版本，但很长一段时间我都没有自己的账户，即使在那时，它也只是我伴侣父母订阅的用户资料。这种观看内容的方式是断断续续且临时的。首先，我必须知道自己想看什么，这意味着我会听从朋友或社交媒体或博客上推荐的节目。然后我必须自己去找，设置好电脑，在昏暗的房间里，沐浴在附近屏幕的冷光中。这与上世纪90年代我小时候打开客厅电视，随便看有线电视节目的日子截然不同。

For most of my adult life, I didn’t own a television. In college, I usually pirated the shows I wanted to watch, like Mad Men, and the illicit habit extended into the Game of Thrones era in the 2010s. I consumed most TV shows and films via the classic setup of a laptop propped up next to me in bed, my head tilted up by pillows to achieve the perfect viewing angle. Netflix launched the earliest version of its streaming service in 2007, but I didn’t have my own account for a long time, and even then it was a user profile on my partner’s parents’ subscription. This mode of viewing content was intermittent and ad hoc. First, I had to know what I wanted to watch, which meant following suggestions from friends or recommendations found on social media or blogs. Then I had to seek it out and set up my computer, bathing in the cold light of the close-by screen in a darkened room. It was a long way from turning on the living room television as a child in the 1990s and watching whatever was on cable.

但在2020年新冠疫情爆发期间，我的观影习惯发生了改变。随着隔离消息的传出，以及全球供应链放缓的迹象开始显现，我唯一确定要为公寓添置的物件就是一台大电视，这是我给自己买的第一台电视——一台60英寸的平板电视，像个外星人一样矗立在客厅里，通体漆黑亮丽，塑料质感十足。它取代了角落里的一把椅子。有一段时间，我和我的伴侣杰西无法适应它在我们公寓里的存在，但后来我们别无选择：在封锁期间，除了追剧之外，我们没什么事可做。所以我最终注册了自己的Netflix账户。

But my viewing habits changed during the onset of the COVID-19 pandemic in 2020. As news of quarantine set in and hints of global supply-chain slowdowns started, the one thing I made sure to acquire for my apartment was a large television, the first one I had ever bought for myself—a sixty-inch flat-screen that loomed in the living room like an alien, all jet-black shiny plastic. It replaced a chair in the corner. For a while, my partner, Jess, and I couldn’t get used to its presence in our apartment, but then we had no choice: during lockdown there wasn’t much to do besides catch up on the backlist of prestige TV shows. So I finally started my own Netflix account.

Netflix 的主页上滚动着网格状的缩略图，用于宣传每部独立的剧集或电影，就像 Spotify、Instagram 或 Twitter 一样，成为了另一个内容推送平台。那些日子里，我频繁地查看主页，这让我更加清楚地意识到它的界面是如何根据我的喜好进行调整的。它逐渐将特定类别的剧集移到我的主页顶部：旅行纪录片、烹饪节目和国际悬疑迷你剧。这些精选被归类在“凯尔的精选”和“为你推荐的类别”等标语下，承诺提供个性化服务。无论是从上到下的分类行，还是从左到右的单个剧集，都是通过算法排序的。正如 Netflix 在其官方帮助中心中所描述的那样：“我们的系统对剧集的排序方式旨在以最佳顺序呈现您可能喜欢的剧集。”

The Netflix home page, with its scrolling grid of thumbnails advertising each separate show or film, became another content feed, like Spotify, Instagram, or Twitter. How often I checked it in those days made me much more aware of how the interface was tailoring itself to what it perceived to be my preferences. It gradually brought particular categories of shows to the top of my home page: travel documentaries, cooking shows, and international mystery miniseries. These selections were gathered under taglines like “Top Picks for Kyle” and “Categories for You,” promising personalization. Both the categorized rows, from top to bottom, and the individual shows, from left to right, are algorithmically ordered. As Netflix describes it in its official Help Center, “Our systems have ranked titles in a way that is designed to present the best possible ordering of titles that you may enjoy.”

Netflix 率先通过推荐引擎进行文化过滤。在 2007 年推出流媒体服务之前，Netflix 还只是一个提供 DVD 邮购租赁服务的系统，它在其网站上推出了 Cinematch 模块，可以根据其他用户的评分（满分五星）为用户推荐电影。这是一种社交信息过滤形式，与上一章提到的早期音乐推荐系统 Ringo 相差无几。Cinematch 推出了2002 年。多年来，事实证明，这些预测在四分之三的时间内准确度在半星以内，Netflix 用户租借了 Cinematch 推荐的电影后，有一半给了它五星评价。2006 年，Netflix 为机器学习工程师举办了一场比赛，要求他们将推荐提高 10%，即构建更好的算法，以赢得 100 万美元的奖金。2008 年，一位参赛开发人员告诉《纽约时报》，他在预测一组特定电影时遇到了困难，这个问题一旦解决，他的算法得分就可以提高 15%。这些电影是《大人物拿破仑》、 《杯酒人生 》、 《 迷失东京》和《杀死比尔：第一卷》等——故事情节离奇，美学风格两极分化，观众要么非常喜欢，要么非常讨厌。这些电影可能属于“邪典电影”之类的类别，尽管它们具有重要的文化意义，但其吸引力并不能完全用数学来概括。这是算法规范化仍然无法实现的品质，因此在 Filterworld 时代往往会缺失。

Netflix pioneered the filtering of culture through recommendation engines. Before it debuted its streaming service in 2007, when it was still just a system of mail-order rental DVDs, Netflix had Cinematch, a module on its website that recommended movies for users, based on other users’ ratings (out of five stars), a form of social information filtering not far from Ringo, the early music recommendation system mentioned in the previous chapter. Cinematch launched in 2002. Over the years, the predictions proved to be accurate within half a star three-quarters of the time, and half of Netflix users who rented a movie that Cinematch recommended rated it five stars. In 2006, Netflix created a contest for machine-learning engineers to improve on the recommendations by 10 percent—building a better algorithm—for a chance at a $1 million prize. One developer in the contest told The New York Times in 2008 that he was having trouble with a specific set of movies, a problem that, once solved, could improve his algorithm’s score by as much as 15 percent. They were films like Napoleon Dynamite, Sideways, Lost in Translation, and Kill Bill: Volume 1—quirky stories with polarizing aesthetics that a viewer could either really like or really hate. These films might fall into a category like “cult classics,” their appeal not quite reducible into math despite their cultural importance. It’s a quality that algorithmic normalization still works against, and thus tends to be missing in the Filterworld era.

2009年，由AT&T研究工程师领导的BellKor in Chaos团队与Pragmatic Theory团队合作，赢得了Netflix的竞赛。他们共同开发了一个名为Pragmatic Chaos的工具，其准确率比原始算法高出10.06%。其中一项重大创新是融入了“奇异值分解”，这是一种算法策略，将具有特定特征的电影（例如爱情片或喜剧片）归类。增加奇异值分解层级可以对更细微的因素进行分类，例如不包含血腥场面的动作片。品味的概念变成了一系列越来越精细的偏好，例如喜欢A而不是B，而不是更深层次、更全面的自我感知。

In 2009, a team called BellKor in Chaos, led by AT&T Research engineers, along with another team called Pragmatic Theory, won Netflix’s competition. Together they created a tool called Pragmatic Chaos that beat the original algorithm by 10.06 percent. One major innovation was incorporating “singular value decomposition,” an algorithmic strategy grouping together movies that shared specific qualities, like romances or comedies. Added layers of singular value decomposition could sort for subtler factors, like action movies that don’t feature gore. The idea of taste became a series of ever more granular preferences, liking A instead of B, rather than a deeper-seated, holistic sense of self.

Cinematch 在 Netflix 网站顶部推出的“你会喜欢的电影”缩略图，是如今 Netflix 观众熟知并喜爱（或厌恶）的更具动态性的流媒体主页的前身。Spotify 的推荐会随着每首歌曲的结束而快速且反复地出现，而电视流媒体服务的算法反应则较慢。用户选择观看新剧集的频率要低得多，而且大多数时候平台只需要播放剧集的下一集，这种推荐无需任何计算。但 Netflix 的功能与 Spotify 类似，引导我们对特定内容的选择，并塑造我们的感知。作为某一特定流派的代表，就像 Galaxie 500 的“Strange”误导性地代表了该乐队的音乐一样。

Cinematch’s line of “movies you’ll love” thumbnails at the top of the Netflix website was a predecessor to the much more dynamic streaming home page Netflix viewers know and love (or hate) today. While Spotify’s recommendations happen quickly and repeatedly, with the end of each song, the algorithms of TV streaming services are slower to act. The user chooses a new show to watch far less often, and most of the time the platform needs only to play the next episode of a series, a recommendation that requires no calculation. But Netflix nevertheless functions similarly to Spotify, guiding our choices toward particular content and shaping what we perceive as representative of a given genre, the same way that Galaxie 500’s “Strange” became misleadingly representative of the band’s music.

Netflix 应用程序内的搜索功能速度慢且不准确；很难按类型搜索，也无法根据演员或导演等信息进行筛选。（这种缺陷催生了网上发布的一系列搜索引擎优化文章，列出 Netflix 上的内容和未提供的内容，就像流媒体的电话簿一样。）搜索结果还会受到其他用户之前操​​作的影响——它不仅仅是一个信息索引，还是另一种推荐信息流，更有可能显示其他人已经喜欢的内容。Netflix 几乎完全抑制了用户的观看意图。主页成为了主要的发现途径，影响着观众的观看内容和时间。根据卡洛斯·A·戈麦斯-乌里韦和尼尔·亨特在 2015 年进行的一项研究，超过 80% 的用户流媒体时间是由 Netflix 的推荐引擎驱动的。在该公司 2018 年以 Netflix Research 名义发布的一段宣传视频中，一位名叫艾什·芬顿的机器学习经理表示：“我们所做的几乎一切都是推荐算法。”

The search function within Netflix’s app is slow and inexact; it’s difficult to search by genres and impossible to filter by information like actors or directors. (This lack has given rise to an entire genre of search-engine-optimized articles published online that list what is and isn’t on Netflix, like a telephone book for streaming.) The search results are also influenced by the previous actions of other users—it’s not just an index of information but another recommendation feed, more likely to surface what other people already like. Users’ intentions are all but discouraged by Netflix. The home page becomes the main method of discovery, influencing what viewers watch and when. More than 80 percent of streaming time from users was driven by Netflix’s recommendation engine, according to a 2015 study by Carlos A. Gomez-Uribe and Neil Hunt. In a publicity video published by the company under Netflix Research in 2018, a machine-learning manager named Aish Fenton states: “Pretty much everything we do is a recommendation algorithm.”

Netflix 的算法会考虑用户的观看历史和评分；其他具有相似偏好的用户的行为；以及内容本身的信息，例如类型、演员和发布日期。此外，算法还会考虑用户观看的时间、观看设备以及他们在该场景下的观看时长。Netflix 表示，年龄、种族和性别等特定用户人口统计数据不被考虑在内——这些变量可能被认为与偏见有关——但这些身份信息通常可以从用户的其他信息中推断出来。Netflix 最终将基于内容的过滤和协同过滤结合在一起。主页变成了一面魔镜，无需用户输入即可呈现他们当时可能想要观看的内容。它消除了早期数字文化时代必须进行的选择的负担和更具意识的选择过程。与 Filterworld 的许多方面一样，算法界面呈现为一个中立的渠道、一个开放的窗口，或者更具体地说，是一个对你个人品味的准确反映和强化。然而，它远非中立。

The Netflix algorithm factors in a user’s viewing history and ratings; the actions of other users with similar preferences; and information about the content itself, like genre, actors, and release dates. It also includes the time of day the user is watching, what device they’re watching on, and how long they tend to watch in that context. Netflix states that specific user demographics like age, ethnicity, and gender are not factored in—those variables might be perceived as adjacent to bias—but such identities can often be implied from other information about a user. Netflix ultimately combines content-based filtering and collaborative filtering. The home page becomes a magic mirror, requiring no input from the user to present what they might want to consume in that moment. It removes the burden of choice and the more intentional process of selection that had to happen in earlier eras of digital culture. Like so many aspects of Filterworld, the algorithmic interface presents itself as a neutral conduit, an open window or, more specifically, an accurate reflection and intensification of your personal taste. Yet it is far from neutral.

2021 年，一位名叫 Niko Pajkovic 的传播策略师在新媒体上发表了一篇关于 Netflix 推荐系统的论文期刊Convergence 上发表的文章《算法正在取代人类的文化意义和决策过程》（至少是数字化程度较低的过程）写道。帕伊科维奇着手测试 Netflix 对个人品味的影响，为此设计了一组具有不同典型人格的虚拟账户。这些典型人格从“铁杆体育迷”到“文化势利者”再到“无可救药的浪漫主义者”不等。体育迷会观看任何与剧烈体育活动有关的电影，无论是虚构的还是纪录片，他们也喜欢超级英雄电影，但鄙视浪漫喜剧。势利者追求晦涩的艺术电影和外国导演的作品，以及任何能够挑战他们情感的东西。但他们不怎么看电视，讨厌真人秀。浪漫主义者喜欢激情和高强度的戏剧，在高雅与低俗之间游走。 Netflix 会要求新用户选择一些符合自己口味的内容，以便初步形成推荐，但帕伊科维奇什么都没选，不同账户的主页一开始基本一模一样。但这种情况很快就发生了变化。

In 2021, a communication strategist named Niko Pajkovic published a paper on Netflix’s recommender system in the new-media journal Convergence. “Algorithms are replacing the fundamentally human—or at least, the less digitally mediated—process of cultural meaning—and decision-making,” he wrote. Pajkovic set out to test Netflix’s impact on personal taste, and to do so designed a set of fake accounts with different archetypal personalities. The personalities ranged from “The Die-Hard Sports Fan” to “The Culture Snob” and “The Hopeless Romantic.” The sports fan watched anything to do with strenuous physical activity, whether fictional or documentary, enjoyed superhero movies on the side but disdained romantic comedies. The snob pursued obscure art-house films and foreign directors, anything that would challenge their sensibilities. But they didn’t watch much TV and hated reality shows. The romantic liked passion and high drama, moving between high- and lowbrow. Netflix asks new users to select a few things that match with their tastes, to shape initial recommendations, but Pajkovic didn’t select anything, and the respective home pages for the different accounts started off, for the most part, identical. But that quickly changed.

Pajkovic 依次登录每个账户，观看符合角色小说品味的节目和电影，并随机选择观看时间以消除变量。观看第二天起，个性化设置就开始了，账户主页开始出现差异。到第五天，浪漫主义者的账户有一排广告“适合无可救药的浪漫主义者的电影”，而势利者的账户有一排广告“广受好评的作者电影”。更通用的主页类别，如“熟悉的收藏”和“精彩的电影”，也包含个性化结果，反映的是同一种内容：浪漫主义者观看浪漫喜剧，体育迷观看运动员纪录片。也许这种略带误导的策略对消费者来说是一种享受：你并不总是希望主页上的文字标签直白地提醒你，你观看的都是浪漫喜剧。

Pajkovic logged in to each of the accounts in turn and watched the shows and films that were appropriate to the character’s fictional taste, randomizing the time of day to eliminate that variable. As soon as the second day of viewing, personalization had started, and the accounts’ home pages began to diverge. By day five, the romantic’s account had a row advertising “Movies for Hopeless Romantics” and the snob’s a row of “Critically Acclaimed Auteur Cinema.” More generic home-page categories like “Familiar Favorites” and “Exciting Movies” also held personalized results, reflecting the same kind of content: romantic comedies for the romantic and athlete documentaries for the sports fan. Perhaps this strategy of slight misdirection is flattering to the consumer: You don’t always want to be blatantly reminded by the text labels on the home page that all you watch are romantic comedies.

Netflix 的算法会将用户划分到特定的“品味社群”，目前这类社群有两千多个。此外，还有七万七千多个“另类类型”或小众类别，包括“理性的法国艺术电影”、“20 世纪 70 年代非裔美国动作冒险片”以及“基于现实生活的感人战争剧”。用户通常不知道这些分类是如何划分的。他们已经被分类了，而且不了解这些不同细分市场的全部内容。用户并非有意选择这些类别；他们只是看到了算法的结果，主页会缩小到他们最有可能开始观看的主题。

The Netflix algorithm slots users into particular “taste communities,” of which there are more than two thousand. And there are more than seventy-seven thousand “altgenres” or niche categories, which include “Cerebral French Art House Movies,” “African-American Action & Adventure from the 1970s,” and “Emotional War Dramas Based on Real Life.” Users are often unaware of how they’ve been categorized, as well as unaware of the full array of these various niches. The user doesn’t choose these categories intentionally; they just see the algorithmic results as their home page narrows to themes that they’re most likely to start watching.

比推荐本身更引人注目的是，Netflix 还会通过算法更改所有内容的缩略图，以根据特定用户进行定制。这一技巧始于 2017 年底，当时名为“图片个性化”。Pajkovic 在他的测试账户中观察到了这种情况。在观看两周后，页面上的所有缩略图看起来都很相似：“在为‘无望的浪漫主义者’推荐的前 10 部影片中，分布在前两行，其中 5 部影片的图片包含浪漫的拥抱（例如，一对情侣接吻或凝视对方的眼睛）。”同样，体育迷的主页上也充斥着男性人物的戏剧性动作图片：挥拳、守门、骑公牛。在屏幕截图中，这些图片看起来几乎令人作呕，内容千篇一律，就像餐厅菜单上只有汉堡包一样。

Even more striking than the recommendations themselves, Netflix also algorithmically changes the thumbnail art on all its content to tailor it to the specific user. That trick began in late 2017 under the label of “artwork personalization.” Pajkovic observed it happening to his test accounts. At the end of two weeks of watching, all of the thumbnails on the page looked similar to each other: “Of the first 10 titles suggested for the Hopeless Romantic, which were spread across the top two rows, five of them presented artwork images containing a romantic embrace (e.g., a couple kissing or staring into one another’s eyes).” Similarly, the sports fan’s home page was covered in images of individual male figures performing dramatic actions: throwing a fist, tending a soccer goal, riding a bull. In screenshots, the spreads look almost nauseating, the content homogenous, like a restaurant menu of only hamburgers.

有时，图片也具有误导性。对于 Netflix 制作的剧集《外滩探秘》，体育迷收到的缩略图是两个角色扛着冲浪板走向水域，而浪漫爱好者收到的则是两个角色即将接吻的特写镜头。这两张图片都迎合了账户的特定偏好，从他们的观看历史可以看出。但这两张图片都没有真正代表这部剧，它是一部涉及失踪和谋杀的动作悬疑剧。用户们有理由担心，Netflix 的主页只显示与他们肤色相同的缩略图，尽管理论上它不跟踪他们的种族。2018 年出现了一场争议，一些人发现，向他们推荐浪漫喜剧《真爱至上》时，黑人演员切瓦特·埃加福特的形象很突出，但他在电影中只扮演了一个小角色。通过如此激进地更改节目缩略图，该平台正在操纵用户，它不是推荐他们可能喜欢的内容，而是改变同一内容的呈现方式，使其看起来更符合他们的偏好。算法图像选择正在迫使你观看一些你原本可能不关心的内容——尤其是在它本来诚实地标注——万一它最终真的对你有用呢？这与界面传达的个性化信息背道而驰，也与培养个人品味背道而驰，因为它从一开始就是一种操纵。

At times, the images were also misleading. For the Netflix-produced series Outer Banks, the sports fan received a thumbnail of two characters carrying surfboards to a body of water, while for the romantic it was a close-up of two characters about to kiss. Both hewed toward the accounts’ specific preferences, as evidenced by their watching history. But neither image really represented the show, which was an action-driven mystery dealing with disappearances and murders. Users have observed, with justifiable anxiety, how Netflix’s home page only displays thumbnails with their same skin color, despite theoretically not tracking their race. The year 2018 saw a controversy in which some people observed that the rom-com Love Actually was recommended to them with prominent imagery of the Black actor Chiwetel Ejiofor, who plays only a minor part in the film. By changing the show’s thumbnail in such an aggressive way, the platform is manipulating the users, not recommending what they might like but altering the presentation of the same content to make it appear more similar to their preferences. The algorithmic image selection is pushing you to watch something that you might not have otherwise cared about—particularly if it had been labeled honestly—on the off chance that it does end up working for you. It’s the opposite of the interface’s message of personalization, and the opposite of the cultivation of personal taste, since it’s a manipulation from the start.

Pajkovic 发现了另一种操纵行为，他用一个账户观看了随机选择的内容，作为实验的对照。在其主页上，该账户最终播放了《速度与激情》系列的全部八部电影——事实上，《速度与激情》系列电影被推荐给了 Pajkovic 的每个账户，尽管除了 Netflix 支付了高额许可费之外，并没有什么其他理由。Pajkovic 写道，该算法“默认推荐那些更有可能引发用户参与的内容，并以个性化为幌子”。你可能喜欢《速度与激情》电影，因为很多人普遍喜欢它们，但这与《速度与激情》成为你特定偏好的目标并不相同。归根结底，Netflix 的此类推荐并非旨在找到符合用户偏好的内容，而更多的是呈现已经流行或易于获取的内容，这是一种品味的幻觉。 2023年，Netflix的流媒体电影库存不足4000部，比百视达（Blockbuster）倒闭前库存量还要低，后者通常库存量超过6000部。Netflix的推荐营造出一种多样性和深度的假象，而现实中并不存在这种现象。

Pajkovic discovered another manipulation with an account that he used to watch a random selection of content, as a control for the experiment. On its home page, that account ended up being served all eight films in the Fast & Furious franchise—in fact, Fast & Furious films were recommended to every one of Pajkovic’s accounts, though there wasn’t necessarily a justification beyond the fact that Netflix was paying a high fee to license them. The algorithm “defaulted to recommending content with a high likelihood of producing user engagement and did so under the guise of personalization,” Pajkovic wrote. You might like Fast & Furious films because lots of people like them generally, but that’s not the same as Fast & Furious being a target of your specific preferences. Ultimately, such Netflix recommendations are less about finding the content that suits a user’s preferences and more about presenting what’s already popular or accessible, an illusion of taste. In 2023, Netflix was streaming fewer than four thousand films, a lower total than what one of the larger Blockbuster stores stocked before that company disappeared, often upwards of six thousand films. The recommendations create an illusion of diversity and depth that doesn’t exist in reality.

在“过滤世界”时代，品味的意义被掏空，这与数字平台衡量参与度的方式有着某种共通之处：它是一种仓促的判断，主要基于某件事是否能立即引起你的喜欢或厌恶。品味的道德价值，即它通常能够引导个人走向更美好的社会和文化，正在逐渐消失。相反，品味相当于一种消费主义，你买什么或看什么，最终决定了你的身份，也决定了你未来的消费。

The hollowed-out meaning of taste in the Filterworld era has something in common with the way engagement is measured by digital platforms: it’s a snap judgment predicated mostly on whether something provokes immediate like or dislike. Taste’s moral capacity, the idea that it generally leads an individual toward a better society as well as better culture, is being lost. Instead, taste amounts to a form of consumerism in which what you buy or watch is the last word on your identity and dictates your future consumption as well.

Pajkovic 虚构的 Netflix 用户并非完全正常——很少有消费者会如此专注于特定类别，因此主页可能会更加多样化。但推荐机制和缩略图的定制会削弱我们对文化的感知，并限制我们拓展视野的可能性。在引导我们进入特定类别的过程中通过软强制，Netflix 算法最终将我们的品味定义为一种固定的东西，并在平台上的每一次互动中变得更加僵化，陷入更深的类别划分。即使推荐准确，它们也可能变得局限。正如 Pajkovic 所写，“反馈循环会强化用户已有的偏好，减少他们接触多样化文化的机会，并否认艺术、美学和文化的对抗性社会角色。”这种缺乏对抗性令人担忧。并不是说伟大的艺术必须天生具有冒犯性；而是当一切都符合既定的期望时，我们就会错过真正进步和令人不适的文化，这些文化可能会颠覆而不是完全融入类别。

Pajkovic’s made-up Netflix users aren’t completely normal cases—few consumers focus so closely on a specific category, and so home pages are likely to be more diverse. But the mechanism of recommendations and the tailoring of the thumbnail images serve to flatten our perception of culture and limit the possibilities for expanding our horizons. In guiding us into particular categories through soft coercion, the Netflix algorithm ends up defining our taste as only one fixed thing, made more rigid by every successive interaction on the platform, moving deeper into a pigeonhole. Even when the recommendations are accurate, they can become limiting. As Pajkovic wrote, “Feedback loops reinforce a user’s pre-existing preferences, diminishing their exposure to a diverse range of cultural offerings and denying art, aesthetics and culture of its confrontational societal role.” That lack of confrontation is concerning. It’s not that great art needs to be inherently offensive; rather, when everything conforms to established expectations, we miss out on culture that is truly progressive and uncomfortable, that might subvert categories rather than fit neatly into them.

Spotify 的运作方式与 Netflix 类似，将用户按预先定义的品味进行分类。与其他平台一样，其算法也常常存在缺陷、偏见和不中立。2019 年 9 月，乡村音乐明星玛蒂娜·麦克布莱德 (Martina McBride) 尝试在 Spotify 上创建一个乡村音乐播放列表。该平台可以自动推荐歌曲添加到播放列表中。麦克布莱德发现，该算法连续 14 次只提供男性乡村歌手的歌曲列表，一次又一次地刷新，最终才出现一位女性音乐家。麦克布莱德对此感到震惊，并在 Instagram 上发帖：“这是懒惰的表现吗？这是歧视吗？这是对音乐的麻木吗？这是脱离实际吗？”

Spotify works similarly to Netflix, bracketing users into predefined categories of taste, and like other platforms, its algorithm is most often flawed, biased, and non-neutral. In September 2019, the country music star Martina McBride attempted to create a country music playlist on Spotify. The platform can automatically recommend songs to add to a playlist. McBride discovered that the algorithm offered lists of songs only by male country artists fourteen times in a row, refresh after refresh, before it came up with a single female musician. McBride was shocked, posting on Instagram: “Is it lazy? Is it discriminatory? Is it tone deaf? Is it out of touch?”

渥太华大学研究乡村音乐电台播放的教授贾达·沃森亲自尝试了这项功能，并得到了类似的结果：她刷新了 12 次，每次刷新的都是男性。尽管出于研究目的，沃森仅使用 Spotify 收听女性歌手的歌曲，但她发现“在前 200 首歌曲（刷新了 19 次）中，只有 6 首（3%）女性歌手的歌曲和 5 首（3%）男女合唱团的歌曲被收录（所有歌曲都在 121 首男性歌手的歌曲之后出现）。”这些推荐显然与她在 Spotify 上听的其他内容没有太大关系。麦克布莱德发现，播放列表推荐功能根本不是基于用户的收听习惯，而是基于播放列表的标题。因此，根据 Spotify 的算法，人们可能会认为乡村音乐就等同于男性，即使沃森创建了一个名为“阴道音乐家的乡村音乐”的播放列表，这个公式仍然成立。有偏见的算法以有偏见的方式定义了这一流派。这是“非常狭隘的视角“乡村音乐意味着什么，”沃森告诉我。尽管 Spotify 传达了个性化的信息，但它却在这个案例中造成了多样性的真空。

Jada Watson, a professor at the University of Ottawa who studies country radio airplay, tried the feature for herself and ended up with a similar result: She went through twelve refreshes of all men. Even though, for research purposes, Watson uses Spotify solely to listen to women artists, she found that “within the first 200 songs (19 refreshes), only 6 songs (3%) by women and 5 (3%) by male-female ensembles were included (all emerging after 121 songs by male artists).” The recommendations clearly didn’t have much to do with what else she listened to on Spotify. McBride discovered that the playlist-recommendation function wasn’t based on the user’s listening habits at all, but on the title of the playlist. So, per Spotify’s algorithm, one might assume that country music is equated with men, a formula that still held even when Watson made a playlist called “Country Music by Musicians with Vaginas.” A biased algorithm defined the genre in a biased way. It’s a “very narrow perspective of what country music means,” Watson told me. Despite its message of personalization, Spotify in this case created a vacuum of diversity.

2014年，一位名叫克里斯蒂安·桑德维格（Christian Sandvig）的学者在微软研究院的一篇博客文章中，创造了“腐败个性化”（corrupt personalization）一词，用来形容诸如Netflix操纵性的电影缩略图和Spotify千篇一律的播放列表等有缺陷的推荐系统。桑德维格写道：“腐败个性化是指你的注意力被吸引到不属于你自身兴趣爱好上的过程。” 推荐系统“服务于一种往往与我们的利益相悖的商业利益。” 以Netflix为例，误导性的图片和无处不在的“速度与激情”推荐或许可以提高用户参与度，让人们认为他们从服务中获得了价值，从而继续续订，维持公司的增长。

In a 2014 post on a blog run by Microsoft Research, a scholar named Christian Sandvig coined the term “corrupt personalization” to describe such flawed recommendations as the manipulative Netflix movie thumbnails and homogenous Spotify playlists. “Corrupt personalization is the process by which your attention is drawn to interests that are not your own,” Sandvig wrote. The recommendation system “serves a commercial interest that is often at odds with our interests.” In the case of Netflix, the misleading images and the omnipresent Fast & Furious suggestions may serve to increase user engagement and encourage people to think they’re getting value out of the service and thus continue renewing their subscriptions, maintaining the company’s growth.

其他腐败个性化的例子包括亚马逊在其市场中优先推荐自有品牌，以及谷歌搜索将公司的其他产品（例如谷歌地图）作为最佳信息来源。公司获利，但用户可能受损，而且这会破坏整体文化生态系统。正如桑德维格所写：“随着时间的推移，如果人们经常被提供与他们的兴趣不符的东西，他们就会被教导想要什么……他们可能会错误地认为这些才是他们真正的兴趣所在，并且很难以其他方式看待世界。” 互联网越来越被封闭在一系列泡沫中，这些泡沫和自我强化的空间使得人们越来越难以找到多元化的视角。这种想法在政治领域很常见——自由主义者主要消费反映其信仰的数字内容，保守主义者也是如此——但它也适用于文化领域。就个人品味而言，了解自己喜欢什么很难，但当某样东西被如此大力地宣传为“为你”时，同样难以知道自己何时不喜欢或不想要。在 Filterworld 中，越来越难以相信自己或知道在算法推荐的认知中“你”是谁。

Other examples of corrupt personalization include Amazon suggesting its own in-house brands before other results in its marketplace and Google Search prioritizing the company’s other products, like Google Maps, as the best sources of information. The company profits, but the user may suffer, and it degrades the overall cultural ecosystem. As Sandvig wrote: “Over time, if people are offered things that are not aligned with their interests often enough, they can be taught what to want…. They may come to wrongly believe that these are their authentic interests, and it may be difficult to see the world any other way.” The Internet has been increasingly enclosed into a series of bubbles, self-reinforcing spaces in which it becomes harder to find a diverse range of perspectives. This idea is familiar from politics—liberals mainly consume digital content that reflects their beliefs, as do conservatives—but it applies to culture as well. In the matter of personal taste, knowing what you like is difficult, but it’s equally hard to know when you don’t like or don’t want something when it’s being so strenuously presented as “For You.” In Filterworld, it becomes increasingly difficult to trust yourself or know who “you” are in the perceptions of algorithmic recommendations.

2011 年，作家兼互联网活动家 Eli Pariser 出版了他的书《过滤泡沫》，描述了算法推荐和其他数字通信途径如何将互联网用户隔离在只会遇到与自身意识形态相符的人。自那时起，过滤泡沫的概念在过去十年中一直备受争议，尤其是在政治新闻媒体的语境下。一些评估，例如阿克塞尔·布伦斯（Axel Bruns）2019 年出版的《过滤泡沫真实存在吗？》（Are Filter Bubbles Real?），得出结论认为其效果有限。其他科学研究，例如 2016 年《舆论季刊》（Public Opinion Quarterly）对过滤泡沫的调查，发现存在一定程度的“意识形态隔离”，尤其是在观点驱动的内容方面。

In 2011, the writer and Internet activist Eli Pariser published his book The Filter Bubble, describing how algorithmic recommendations and other digital communication routes can silo Internet users into encountering only ideologies that match their own. The concept of filter bubbles has been debated over the decade since then, particularly in the context of political news media. Some evaluations, like Axel Bruns’s 2019 book Are Filter Bubbles Real?, have concluded that their effects are limited. Other scientific studies, like a 2016 investigation of filter bubbles in Public Opinion Quarterly, found that there is a degree of “ideological segregation,” particularly when it comes to opinion-driven content.

然而，文化和文化品味与网络上的政治内容和意识形态信仰有着不同的动态；即使它们在相同的信息流中传播，驱动它们的动机也不同。政治过滤气泡会因分歧而将用户分成对立的派系，而文化推荐则将他们聚集在一起，共同构建越来越大的受众群体，以吸引那些共同点最不重要的内容。算法文化聚集在中心，因为消费某种文化的决定很少出于仇恨或冲突。印第安纳大学研究推荐算法的张菁菁教授在合作开展一项关于个性化音乐推荐的实验时，发现了同质化理论的证据，该实验在2012年都柏林推荐系统会议上进行了展示。

Yet culture and cultural taste have different dynamics than political content and ideological beliefs online; even though they travel through the same feeds, they are driven by different incentives. While political filter bubbles silo users into opposing factions by disagreement, cultural recommendations bring them together toward the goal of building larger and larger audiences for the lowest-common-denominator material. Algorithmic culture congregates in the center, because the decision to consume a piece of culture is rarely motivated by hate or conflict. Jingjing Zhang, a professor at Indiana University who studies recommendation algorithms, found evidence for the theory of homogenization when she collaborated on an experiment about personalized music recommendations, presented at the 2012 Dublin Conference on Recommender Systems.

学生们被展示了一些歌曲，他们被告知这些歌曲是根据他们的个人品味推荐给他们的，并以星级评分来表示——但这些评分实际上是人为的、任意的。然后，学生们被问及他们愿意为一首特定歌曲支付多少钱；星级越高，他们愿意支付的价格就越高。每增加一颗星，他们为这首歌曲支付的意愿就会增加10%到15%。实验表明，对推荐的感知会扭曲特定文化的感知价值，使其看起来更受喜爱或更具意义。由于算法推荐的自我强化循环，这一缺陷愈发严重。正如张在播客《星球金钱》（Planet Money）中所说，随着时间的推移，系统将“提供更少的多样性推荐”。最终，她说，它将“向每个人提供类似的内容，无论个人品味如何。”这就是我们今天所经历的同质化。

Students were presented with songs that they were told had been recommended to them based on their personal taste, as indicated by a star rating—but the ratings were actually artificial and arbitrary. Students were then asked how much they would pay for a given song; the higher the star rating, the more they were willing to pay. Each added star resulted in a 10 to 15 percent increase in willingness to pay for the song. The experiment showed how the perception of recommendation could skew the perceived value of a given piece of culture, making it seem more likable or significant. The flaw intensifies due to the self-reinforcing loop of algorithmic recommendations. Over time, the system will “provide less diverse recommendations,” as Zhang told the podcast Planet Money. Eventually, she said, it will “provide similar items to everybody, like, regardless of personal taste.” Hence the homogenization we are experiencing today.

收藏文化

COLLECTING CULTURE

在决定我们消费的文化以及我们对它的感受方面，呈现推荐外观的应用界面与实际推荐同样重要。界面设计属于科技行业的“用户体验”标签：用户在导航、搜索和点击时发生的微交互。当今平台的用户体验往往极其被动。你不应该过多地关注底层内容，只需消费已经呈现在你面前的内容即可——理论上，算法比你更了解你自己，但这显然是不正确的。如果我们总是依赖Netflix主页、Instagram的“发现”页面或TikTok的“为你推荐”推送来展示我们感兴趣的内容，那么我们就没有动力去自己决定要寻找什么、关注什么，以及或许最重要的是要保存什么。我们通常通过保存文化碎片来构建个人品味：慢慢地收集对我们重要的东西，就像鸟儿筑巢一样，为我们的喜好树立一座纪念碑。

App interfaces that present the appearance of recommendations can be as important as actual recommendations in determining what culture we consume and how we feel about it. Interface design falls under the tech industry label of “user experience”: the micro-interactions that happen as a user navigates, searches, and clicks. The user experience of today’s platforms tends to be overwhelmingly passive. You’re not supposed to look under the hood too much, just consume what’s already in front of you—theoretically, the algorithm knows you better than you know yourself, though that’s patently untrue. If we can always rely on the Netflix home page, Instagram’s Discover page, or the TikTok “For You” feed to show us something that we’re interested in, then we have less impetus to decide for ourselves what to look for, follow, and, perhaps most important, save. We often build our senses of personal taste by saving pieces of culture: slowly building a collection of what matters to us, a monument to our preferences, like a bird constructing a nest.

但算法推送越自动化，我们作为消费者就越被动，也就越不需要去建立收藏，去保存那些对我们重要的东西。我们放弃了收藏的责任。过去二十年，文化收藏——无论是DVD电影、黑胶唱片还是书架上的书籍——已经从一种必需品变成了一种奢侈的享受。数字平台宣称他们能够随时随地提供一切所需，我又何必操心手头上能找到什么呢？问题在于，数字平台提供的内容并不能保证其持久性——看似完整，实则虚幻，背后是推荐——而且它们的界面也在不断变化。突然改变的界面带来的困惑，在Filterworld里是司空见惯的。

But the more automated an algorithmic feed is, the more passive it makes us as consumers, and the less need we feel to build a collection, to preserve what matters to us. We give up the responsibility of collecting. Over the past two decades, the collecting of culture—whether films on DVD, albums on vinyl, or books on a shelf—has shifted from being a necessity to appearing as an indulgent luxury. Why would I bother worrying about what I have access to at hand when digital platforms advertise their ability to provide access to everything, forever, whenever I want? The problem is that there is no guarantee of permanence in what digital platforms offer—the appearance of totality is a facade, buttressed by recommendations—and their interfaces are constantly changing. The confusion induced by a suddenly changed interface is a common experience in Filterworld.

2021年末的一个早晨，我在笔记本电脑上打开Spotify，突然发现自己迷失了方向。我习惯于通过一组特定的点击来访问我喜欢的音乐——这次是尤瑟夫·拉蒂夫1961年的爵士乐专辑《东方之声》。疫情期间，许多工作日的早晨我都在家里工作，我会把这张专辑这让 Jess 很恼火，她经常嘲笑它重复、不和谐的开头。但那天我找不到那张专辑，也找不到我用 Spotify 点赞按钮保存的完整专辑列表，而这个按钮是该软件将它们集中在一个地方的主要方式。我的肌肉记忆不起作用了。合集被重新排列了，我没有任何通知或选择。感觉就像失语症一样，就好像有人在一夜之间移动了我客厅里的所有家具，而我还像往常一样试图浏览它们。Spotify 阴郁的黑绿色界面中出现了一个新的“你的音乐库”标签，提示了我想要查找的所有内容，但它却打开了一个我不认识的自动生成的播放列表窗口。上面的下一个标签提供了播客，我从未在应用上听过。一切都没有意义。

One morning in late 2021, I opened Spotify on my laptop and found myself suddenly lost. I was used to performing a specific set of clicks to access the music I like—in this instance, a 1961 jazz album by Yusef Lateef called Eastern Sounds. Many weekday mornings working from home during the pandemic I would put the album on first thing, much to the annoyance of Jess, who often laughed at its repetitive, discordant opening notes. But that day I couldn’t find the album, or the full list of albums that I had saved with the Spotify Like button, which was the software’s principal way of keeping them in one place. My muscle memory didn’t work. The collection had been rearranged without giving me any notice or choice in the matter. It felt like a form of aphasia, as if someone had moved around all the furniture in my living room overnight and I was still trying to navigate it as I always had. A new “Your Library” tab in Spotify’s moody black-and-green interface hinted at everything I was trying to find, but instead it opened a window of automatically generated playlists that I didn’t recognize. The next tab over offered podcasts, which I never listened to on the app. Nothing made sense.

随着各种媒体形式纷纷转向流媒体，一切似乎都触手可及，我们很容易忘记，我们也可以与我们在个人时间里消费的文化作品建立一种物理的、非算法的关系。我们把书放在书架上，在客厅墙上挂艺术品，还收藏着成堆的黑胶唱片。当我们想要体验某种东西时，我们会去寻找它：抓住书脊找到一本书，或者从盒子里抽出一张专辑。我们与事物互动的方式以及我们存储它的位置也会改变我们消费它的方式，正如Spotify的更新强行提醒我的那样。同样的事情也发生在Twitter上，它添加了自己的算法“为你推荐”信息流，Instagram也移动了发布照片的按钮，一度用观看TikTok风格视频的按钮取而代之。

As all forms of media have moved into streaming, when everything seems to be a single click away, it’s easy to forget that we can also have physical, non-algorithmic relationships with the pieces of culture we consume in our personal time. We store books on bookshelves, mount art on our living room walls, and keep stacks of vinyl records. When we want to experience something, we seek it out: finding a book by its spine or pulling an album from its case. The way we interact with something, and where we store it, also change the way we consume it, as Spotify’s update forcibly reminded me. The same thing happened with Twitter when it added its own algorithmic “For You” feed and Instagram when it moved the button to post a photo, at one point replacing it with the button to watch TikTok-style videos.

所有这些变化让我渴望一种相反的东西：一种固定、稳定、可靠的方式，去接触任何你想接触的文化。而这正是早期文化收藏和消费形式曾经提供的。我们视这种稳定性为理所当然。1931年，德国文化评论家瓦尔特·本雅明写了一篇名为《打开我的图书馆》的文章，描述了我们与实体文化对象的关系。在文章中，本雅明讲述了他如何将藏书从布满灰尘的板条箱中取出，这些书卷已经装在里面好几年了。这些书卷散落在地板上，“尚未感受到秩序带来的那种轻微的无聊”，它们正准备再次被重新摆放到书架上。对本雅明来说，拥有这些书本身就构成了他的他作为读者、作家和人的身份认同——即使他并未读完所有书籍。这些书骄傲地摆放在他的书架上，象征着他仍然渴望获得的知识，或是他曾经游历过的城市。收藏书籍是他与世界互动的方式，也是他构建世界观的方式，并在批判性写作中不断深化。

All these changes make me crave the opposite: a fixed, stable, reliable way of accessing whatever culture you want. Which is exactly what earlier forms of collecting and consuming culture once offered. We took that stability for granted. In 1931, the German cultural critic Walter Benjamin wrote an essay called “Unpacking My Library,” describing our relationship with physical cultural objects. In the essay, Benjamin narrates removing his book collection from dusty crates, where the volumes had been enclosed for years. The volumes are splayed loose on the floor, “not yet touched by the mild boredom of order,” all set to be rearranged on shelves once more. For Benjamin, the very possession of these books formed his identity as a reader, writer, and human being—even if he hadn’t read all of them. They sat proudly on his shelves as symbols, representing the knowledge that he still aspired to gain or the cities to which he had traveled. Accumulating books was his way of interacting with the world, of building a worldview that he furthered in his critical writing.

本雅明的图书馆是一座个人纪念碑，就像我们用喜欢或认同的物品建造的一样，构建着我们的品味。它的重要性在于它的永恒——藏品是由我们拥有的物品组成的，除非我们决定它们应该消失，否则它们不会消失。“所有权是一个人与物品之间最亲密的关系，”本雅明写道。“并非物品在他身上活了过来；而是他活在物品里。”换句话说，我们常常在身边的物品中发现，甚至重新发现自我。但如果本雅明的书架顺序和藏书目录每隔几个月就变化一次，藏品和个人之间的这种相互依存或共同进化就不会发生。Spotify 的界面更新和算法变化给我的感觉就是这样：彻底颠覆了塑造我的艺术和文化。

Benjamin’s library was a personal monument, the same kind that we all construct of things we like or identify with, building our sense of taste. Its importance lay in its permanence—collections are made up of things that we own, that don’t go away unless we decide they should. “Ownership is the most intimate relationship that one can have to objects,” Benjamin wrote. “Not that they have come alive in him; it is he who lives in them.” In other words, we often discover, and even rediscover, ourselves in what we keep around us. But that codependence or co-evolution of collection and person wouldn’t happen if the order of Benjamin’s shelves and the catalog of his books kept changing every few months. That’s what Spotify’s interface updates and algorithmic changes felt like to me: a total disruption of the pieces of art and culture that shaped me.

在“过滤世界”（Filterworld）中，我们的文化收藏不再完全属于我们。书架仿佛开始实时自动变换形态，将一些内容移到最前面，而其他内容则被淡化。这让我想起一个戏法娴熟的魔术师，他潜意识里鼓励你抽取一张特定的牌——即使魔术师让你相信这是你自己的选择。这种自主性的缺失正在破坏我们与所热爱的文化之间的联系。我们通常不会将书架本身与其中的内容区分开来，但它们确实是很棒的设备。它们展示书籍或专辑，你可以以相对中立的方式从展示的选项中进行选择。只有收藏家才能决定如何安排他们的藏品，可以按作者、书名、主题，甚至封面颜色来排列书籍——而且它们会一直摆放在原处。而我们的数字文化界面则并非如此，它们会遵循拥有它们的科技公司的喜好和优先级。例如，如果 Spotify 突然将播客类别放在一个显眼的新位置，那是因为该公司决定播客将为其带来更多收入未来。界面遵循公司的激励措施，首先推销自己的产品，或者改变熟悉的模式来引导用户尝试新功能。

In Filterworld, our cultural collections are not wholly our own anymore. It’s as if the bookshelves have started changing form on their own in real time, shuffling some material to the front and downplaying the rest. It reminds me of a sleight-of-hand magician who subliminally encourages you pick a specific card—even as the magician lets you believe it’s your own choice. And this lack of agency is undermining our connections to the culture that we love. We don’t often think about bookshelves on their own, separate from what they contain, but they are great devices. They display books or albums, and you can choose from among the displayed options in a relatively neutral way. The collector is the only one who decides how to arrange their possessions, ordering books by author, title, theme, or even color of the cover—and they stay in the same places they’re put. That’s not true of our digital cultural interfaces, which follow the whims and priorities of the technology companies that own them. If Spotify suddenly gives the category of podcasts a prominent new placement, for example, it’s because the company has decided that podcasts are going to make up more of its revenue in the future. The interfaces follow the company’s incentives, pushing its own products first and foremost, or changing familiar patterns to manipulate users into trying a new feature.

本杰明写道，收藏家对自己的藏品有一种“责任感”。但我们很难对自己在互联网上收藏的藏品产生这种归属感；我们无法像本杰明那样，成为我们所欣赏的文化的守护者。我们实际上并不拥有它，也无法保证每次都能以同样的方式访问它。

Benjamin wrote that collectors have a “feeling of responsibility” to their collections. But it’s very difficult to feel such ownership for what we collect on the Internet; we can’t be stewards of the culture we appreciate in the same way as Benjamin. We don’t actually own it and can’t guarantee accessing it in the same way each time.

你可能积累了一个辛苦整理的数字音乐库，但当应用程序发生变化时，它却变得一片混乱。或者，当流媒体服务关闭时，你的收藏可能会完全丢失。数字界面往往会在没有任何警告和记录的情况下发生变化；每次重新设计都会抹去之前的版本。现在不可能像几十年前那样通过拒绝更新来继续使用过时的软件版本，而是访问你最喜欢的旧版 Spotify 或 Instagram。这些应用程序现在大多存在于云端，由用户在线访问，公司完全控制着它们的运行方式。这种不稳定性只会加剧正在发生的文化扁平化，因为用户无法在原来的环境中存储或重温他们过去的体验。存在的只是不断变化的数字现在时。

You can accrue a laboriously curated digital library of music only to have it thrown into disarray when the app changes. Or your collections can be lost entirely when a streaming service shuts down. Digital interfaces tend to change without warning and without record; each redesign erases its previous iterations. There’s no way to access older, favored versions of Spotify or Instagram the way it was possible in decades past to stick with an outdated software version by declining updates. The apps now mostly exist in the cloud, accessed online by the user, and the company wholly controls how they work. This instability only intensifies the cultural flattening going on, since users can’t store or revisit their past experiences within their original context. All that exists is a relentlessly changing digital present tense.

某个应用程序的消失或全面改版，会让之前收集的内容化为乌有。互联网不像一堆八轨磁带，用合适的技术就能再次播放，再次获得完整的体验。在网上建立收藏更像是在沙滩上堆沙堡：最终潮水涌来，它仿佛从未存在过。当我回顾自己在 Tumblr 等老平台上的账户时，我也有同样的感受。在 2010 年代初，我在那里收集了 408 页动漫 GIF、诗歌片段和挽歌般的电子游戏截图；或者，当我回顾自己 2007 年左右在 Facebook 上发布的相册时，我也有同样的感受。如今，相册功能已经逐渐淡出，至少是按照最初的设想。

The disappearance or overhauling of a particular app throws the content gathered there to the wind. The Internet is not like a bunch of eight-track tapes that can be played again with the proper technology and experienced in full once more. Building a collection online more closely resembles building a sandcastle on the beach: eventually the tide washes in, and it’s as if it never existed in the first place. Such is the feeling I get when I look back at my accounts on older platforms like Tumblr, where over the early 2010s I collected 408 pages of anime GIFs, fragments of poems, and elegiac video-game screenshots, or when I scroll back to the photo albums I posted on Facebook around 2007, a feature that has since faded out, at least as it was originally intended.

数字技术的变迁剥夺了我们收藏的意义。它们如今只剩下怀旧的废墟，如同曾经喧嚣都市的残骸般沉寂。我曾经在Tumblr上分享的许多图片现在都成了失效的链接。我本可以我下载了这些收藏的黄金时期，并确保自己随时可以访问它们，但这无法捕捉到它们曾经代表的流动和社交的意义。当我浏览我至今仍保存在网上的 Tumblr 存档时，我看到的是一个比如今算法时代涡轮增压信息流更慢、更私密、更线性、更连贯的数字空间。Tumblr 更像是一个书架。它提醒我曾经的一切都不同，但我并不抱太大希望重新找回那种冥想的节奏。

The shifting sands of digital technology have robbed our collections of their meaning. They appear only as nostalgic ruins, the remains of once-inhabited metropolises gone silent. Many of the images I once shared on Tumblr are now broken links. I could have downloaded these collections in their prime and made sure I could always access them, but that wouldn’t capture the meaning of their flow and the social exchange that they once represented. What I see when I look at my Tumblr archive that’s still extant online is a glimpse of a slower, more intimate, linear, and coherent digital space than what now exists in the turbocharged feeds of the algorithm era. Tumblr was more like a bookshelf. It reminds me that things were once different, but doesn’t give me much hope of recapturing that meditative pace.

十几岁的时候，我把CD音乐装订成册，放在我经常开的家用车里。有些是买的，有些是我自己拼装刻录的，体现了我的个人品味。我现在还留着那个CD，现在看着它——九十年代的橡胶边缘和厚重的布料——让我感到一阵怀旧，也让我回想起里面的音乐。Spotify没有CD活页夹可以随身携带。随着我使用这个平台的时间越来越长，界面也发生了变化，我发现自己变得越来越被动，保存的专辑越来越少，也越来越少去思考它们的主题和故事，把它们看作一个整体。但对我来说，作为收藏家和文化消费者的糟糕之处最终却对这个平台有利：我会继续订阅Spotify，因为这是我获取音乐的唯一途径。

As a teenager, I had a binder of music on CDs that I kept in the family car I usually drove. Some I had purchased; others were mixes that I assembled and burned myself, codifying my personal taste. I still have that binder, and looking at it now—the very nineties rubberized edges and heavy-duty fabric—gives me a sense of nostalgia and a memory of the music contained therein. With Spotify, there’s no CD binder I can take with me. As I’ve used the platform longer, the interface changing shape, I’ve found myself becoming a more passive user, saving fewer albums and thinking less about their themes and stories as cohesive creations. But what’s worse for me as a collector and cultural consumer is ultimately better for the platform: I’ll keep subscribing to Spotify because it’s the only way I’ll have access to my music.

虽然我们拥有选择自由的优势，但算法推送提供的无穷无尽的选项常常会让人产生一种无意义感：我可以听任何东西，为什么任何一件东西对我来说都很重要？收藏与文化之间的建设性关系是双向的。当我们发现一些足够有意义的东西值得保存、值得添加到我们的收藏中时，这种行为不仅会将其更深地铭刻在我们的心中，还会围绕这些物品本身构建一个语境，无论是文本、歌曲、图像还是视频。这个语境不仅是为了我们自己，也是为了其他人，是为了整个文化紧密相连、共享的语境。这就是本杰明所描述的：“收藏现象失去了它的主人，也就失去了意义。” 收藏需要个体的守护者，它们表达着他们的声音和品味。Spotify 的海量收藏并非一个连贯的集合；而是一场雪崩。

While we have the advantage of freedom of choice, the endless array of options presented by algorithmic feeds often instills a sense of meaninglessness: I could be listening to anything, so why should any one thing be important to me? The constructive relationship between collecting and culture goes in both directions. When we find something meaningful enough to save, to add to our collection, the action both etches it a little deeper into our hearts and creates a context around the artifact itself, whether text, song, image, or video. The context is not just for ourselves but for other people, the knit-together, shared context of culture at large. That’s what Benjamin described when he wrote, “The phenomenon of collecting loses its meaning as it loses its personal owner.” Collections need individual caretakers, whose voices and tastes they express. The mass of Spotify isn’t a coherent collection; it’s an avalanche.

用户有时会觉得自己是这场雪崩的无助受害者。流媒体服务分为两种风格正如高管们向我描述的那样，他们利用消费来分析用户。有些时候，用户会“向前倾”（lean-in），专注于选择消费内容，并积极判断结果。有些时候，用户会“后仰”（lean-back），随着内容在后台运行，无需过多关注内容是什么或接下来播放什么。算法推荐将我们推向后者，在这种状态下，我们像鹅肝酱鸭一样被灌输文化，更注重数量而非质量——因为数量，即纯粹的消费时间，才是平台通过定向广告赚钱的关键。

The user can sometimes feel like a hapless victim of that avalanche. Streaming services separate between two styles of consumption to analyze their users, as executives have described to me. There are “lean-in” moments, when the user is paying attention, choosing what to consume and actively judging the result. And there are “lean-back” moments, when users coast along with content running in the background, not worrying too much about what it is or what plays next. Algorithmic recommendations push us toward the latter category, in which we are fed culture like foie-gras ducks, with more regard for volume than quality—because volume, sheer time spent, is what makes money for the platform through targeted advertising.

随着消费者变得越来越被动，无法展现出他们独特的品味，艺术家们被迫更加努力地应对算法的压力，因为通过信息流进行创作是他们获得足够规模的受众和参与度以维持生计的唯一途径。他们需要触达我们当下的处境，而我们目前正深陷信息流，对信息流不太关注，既接受最新的算法推荐，又可能因为一时的不满而转身离开。除了适应，别无选择。

As consumers become increasingly passive, failing to exercise distinctly cultivated tastes, artists are forced to contend even more with algorithmic pressures, because working through the feed is the only way they can reach the scale of audience and engagement that they need to make a living. They need to reach us where we are, and where we are is leaning back in the feed, not paying too much attention, both accepting of the newest algorithmic recommendation and likely to flip away at a moment’s dissatisfaction. There is no choice but to adapt.

创作者遵守 Feed

CREATORS CONFORMING TO THE FEED

TikTok 是一个我加入得晚但又如饥似渴的平台。当时已经是 2020 年了，经历了数月的疫情隔离，我感觉自己已经玩遍了所有可能的在线娱乐方式。但我从未尝试过 TikTok：我觉得它不适合我，因为我是千禧一代，而不是 Z 世代，不属于它的目标受众。但我迫切地想动起来，于是下载了 TikTok 到手机上，打开主页面。我的初始页面立刻就被各种视频片段填满了，看起来似乎是随机的。算法像投掷飞镖一样，瞄准我的个人喜好，看看哪些会落下来——监视我持续观看的内容，以及我通过滑动跳过的内容。一些主题开始浮现：滑板技巧、可爱的狗狗，以及演奏吉他的音乐家。这​​感觉很催眠，尤其是因为我什么都不用做。我只需要向后靠，让我的大脑几乎下意识地决定哪些内容有趣。渐渐地，新的主题会出现在信息流中：旅行见闻、烹饪视频、用简陋工具制作的工艺品荒野。短短几周内，大多数人喜欢的通用类别逐渐精炼成一系列特定的兴趣，我与一小群人分享。这就是算法信息流在起作用，它根据我的喜好将我的偏好分类，然后一遍又一遍地提供这些主题。这是我用过的最个性化、最精准的信息流；因此，它既令人愉悦，又令人恐惧。

TikTok was a platform that I joined late but hungrily. It was well into 2020, and after many months of pandemic quarantine, I felt like I had exhausted all possible forms of online entertainment. But I had never tried TikTok: I thought it wasn’t for me, that as a millennial instead of a member of Gen Z, I wasn’t part of the intended audience. But I was desperate for activity, so I downloaded the app onto my phone and opened it to the main feed. My virgin feed was immediately populated with video clips, seemingly at random. The algorithm was throwing darts at the board of my personal taste to see what landed—surveiling what I kept watching and what I skipped over by swiping. A few themes started to emerge: skateboarding tricks, cute dogs, and musicians playing guitars. It was hypnotic, not least because I didn’t have to do anything. I could just lean back and let my brain almost subconsciously decide what was interesting. Gradually, new themes would emerge in the feed: travelogues, cooking videos, crafts constructed with rudimentary tools in the wilderness. Over just a matter of weeks, a spread of generic categories that most people enjoy refined themselves into a series of specific interests that I shared with a smaller group. This was the algorithmic feed at work, sorting my preferences into categories of taste and then serving up those subjects, over and over. It was the most personalized and accurate feed I had ever used; as such, it felt both pleasurable and horrifying.

TikTok 由一家名为字节跳动的中国公司运营，并拥有中文版抖音，但直到 2018 年才在美国推出。当时字节跳动收购了另一家名为 Musically 的中国社交网络，该社交网络于 2014 年推出，以青少年发布对口型音乐视频而闻名，在美国拥有强大的用户群。Musically 并入了 TikTok，后者因其快速的音乐或舞蹈视频片段而广受欢迎，这是其前身的延续。起初，TikTok 视频最多只有 15 秒，与昙花一现但深受喜爱的视频应用 Vine 相似；然后限制延长到 1 分钟，然后是 10 分钟。TikTok 的另一个与众不同之处在于其主要的“为你推荐”信息流，这几乎完全是算法决定的。用户不被鼓励选择关注谁；他们只是信任程序的决定。（与 Twitter 和 Facebook 等网站不同，它们仍然在某种程度上按时间顺序并基于关注者。）这个噱头奏效了； 2021年，TikTok月活跃用户突破10亿，成为承载我们数字生活的大型社交网络的继承者。它的成功意味着全算法信息流日益成为默认设置，并开始开创一个数字名望和文化成功的新时代。

TikTok is run by a Chinese company called ByteDance and has a Chinese equivalent, Douyin, but it didn’t launch in the United States until 2018. That’s when ByteDance acquired another Chinese social network called Musically, which had launched in 2014 and was best known for teenagers posting lip-synching music videos, with a pre-established strong user base in the United States. Musically merged into TikTok, which became popular for its quick video clips of music or dance, a holdover from its predecessor. At first TikTok videos were just fifteen seconds long at most, an echo of the brief-lived but beloved video app Vine; then the limit was lengthened to a minute, and then ten minutes. What also set TikTok apart was its primary “For You” feed, which is almost entirely algorithmic. Users were not encouraged to choose who to follow; they just trusted the decisions of the equation. (As opposed to the likes of Twitter and Facebook, which remained somewhat chronological and based on follows.) The gimmick worked; in 2021 TikTok passed a billion monthly active users and took its place as a successor in the line of massive social networks that host our digital lives. Its success meant that all-algorithmic feeds increasingly became the default, and it began to create a new era of digital fame and cultural success as well.

我开始在动态里看到一类视频，是一些快速、缺乏叙事的日常生活蒙太奇：倒咖啡的瞬间，整理床铺的瞬间，公寓窗户透进来的光线。这些视频大多是匿名的，创作者拿着手机，拼凑着周围环境的肖像。流行音乐作为背景音乐：耳熟能详的歌曲为这一刻注入了电影般的氛围。这种形式似乎特别适合疫情期间，因为没有其他的奇观，所以不得不将家庭环境“浪漫化”——正如TikTok用户所说。在一则TikTok视频中，一名男子拍摄了自己在一栋高层公寓楼顶层的游泳池里游泳。背景音乐是​​弗兰克·奥申的《白色法拉利》，这是一首柔和的挽歌，讲述的是深夜驾驶的场景。

One category of video I started to see in my feed was quick, narrative-less montages of everyday life: glimpses of coffee being poured, beds being made, light coming through windows of apartments. These were mostly anonymous, the creator holding their phone in front of them and assembling a portrait of their surroundings. Pop music provided the soundtrack: familiar songs that imbued the moment with a cinematic ambience. The format seemed particularly suited to the pandemic, when domestic surroundings had to be “romanticized”—as TikTok users put it—since there were no other spectacles at hand. In one TikTok, a man filmed himself swimming in a lap pool that was on the top floor of a tall apartment building. The soundtrack was Frank Ocean’s “White Ferrari,” a soft, elegiac song about driving late at night.

这就是我对 Nigel Kabvina 作品的介绍，当时他 25 岁，默默无闻，住在英国北部的曼彻斯特。我看到他的游泳视频时，Kabvina 只有几千名粉丝。在接下来的两年里，他的粉丝超过 400 万，跻身 TikTok 创作者的顶级梯队，这要归功于“为你推荐”功能，它能让一段视频瞬间吸引数百万观众，即使这个账号相对默默无闻。（如果推荐已经占据主导地位，就没有必要拥有固定的受众，尽管这并没有什么坏处。）Kabvina 的成功之处在于积极迎合推送，将自己的品味——他的创造性表达——置于算法推广的界限之下。但一开始，他只是在开玩笑。在我们最初的谈话中，他解释说，在制作视频时，他试图捕捉某些感觉，一种 TikTok 特别擅长捕捉的短暂氛围。这是一种抵消隔离压力的方式；卡布维娜在一家连锁特色鸡尾酒吧担任调酒师的工作消失了。

This was my introduction to the work of Nigel Kabvina, who at that point was an unknown twenty-five-year-old living in Manchester, in the north of the United Kingdom. When I saw his swimming video, Kabvina had only a few thousand followers. In the following two years, he would reach over four million followers and enter into the very top echelon of TikTok creators, thanks to the “For You” feed’s ability to rocket a video instantly to millions of viewers, even if the account is relatively obscure. (There’s no need to have an established audience if recommendations are so dominant, though it doesn’t hurt.) Kabvina succeeded by aggressively catering to the feed, subordinating his own taste—his creative expression—to the limits of algorithmic promotion. But in the beginning, he was just messing around. In our first conversations, he explained that in making the videos he was trying to capture certain feelings, an ephemeral atmosphere that the format of a TikTok was particularly good at capturing. It was a way to counteract the stress of quarantine; Kabvina’s bartending job at a chain of specialty cocktail bars had vanished.

卡布维纳也开始在他的公寓里制作烹饪视频，视频里是他制作精美菜肴的短片，通常是为室友准备的早午餐。他在牛油果吐司上刻上巴洛克风格的图案；烤迷迭香，用玻璃杯包裹烟雾；甚至把冰冻成一个盛麦片的水晶碗。我看着他的账号粉丝不断增长，先是几万，然后是几十万。他的视频大受欢迎；视频评论区里，一个活跃的、自带小段笑话的社群逐渐壮大。到2021年8月，卡布维纳的粉丝已达一百万，并决定全职从事TikTok。他很快就获得了谷歌和英国连锁超市森宝利（Sainsbury's）等赞助客户。

Kabvina also began making cooking videos in his apartment, quick clips in which he assembled elaborate dishes, often a brunch for his roommate. He incised baroque patterns into avocado toast; burned rosemary and trapped the smoke under a glass; and even froze ice into a crystalline bowl for cereal. I watched as his account gained followers, first tens of thousands and then hundreds of thousands. His videos were a hit; an active community with its own inside jokes grew in the videos’ comment sections. By August of 2021, Kabvina had reached a million followers and decided to pursue TikTok full-time. He quickly netted sponsorship clients like Google and Sainsbury’s, the British grocery chain.

冬天的一天，我在伦敦亲自见到了卡布维纳，当时他从曼彻斯特出差。我让他选个地方见面，他选了一家在苏荷区有分店的著名鸡尾酒吧Swift——“去伦敦旅行一定要去，”他给我发短信说。我们先在楼上一家铺着地铁瓷砖的店面坐下，然后偷偷溜进了一个卡座。在舒适的地下室里。我采纳了Kabvina的鸡尾酒推荐，先点了一杯爱尔兰咖啡，这是店里的特色菜，非常适合我们下午晚些时候的会面。他穿着一身低调的全黑套装，笑容灿烂而自然。我之所以能认出他的脸，是因为他最近开始在TikTok上露面，而之前他只拍肩膀以下的自己。Kabvina的TikTok粉丝已经足够多，足以支撑他的经济，尽管这庞大的粉丝群仍然让他感到困惑。“想象一下，你走进厨房泡茶，想象有30个人走进房间，目不转睛地盯着你。然后再想象一下，100万，”他说。每个月，他的TikTok视频观看次数都在4000万次。Kabvina的视频、算法推送的信息以及如饥似渴的观众形成了一个反馈循环。他称之为“即时满足”：“我可以在TikTok上发布帖子，10分钟后再查看，就能看到3万人观看了。”

I met Kabvina in person in London one day in winter, while he was on a business trip from Manchester. I asked him to pick a place to meet and he chose Swift, a famed cocktail bar with a branch in Soho—“a must when traveling to London,” he texted me. We first sat upstairs in a subway-tiled storefront and then snuck into a booth in the cozy basement. I took Kabvina’s recommendation on a cocktail to start with, too: an Irish coffee, which was a specialty of the house and ideal for our late-afternoon meeting time. He wore an unobtrusive all-black outfit and had a wide, easy grin. I recognized his face only because he had recently started showing it on TikTok, whereas previously he had filmed himself only from the shoulders down. Kabvina’s TikTok audience had become big enough to financially sustain him, though its vast scale still bemused him. “Imagine going to your kitchen and making a cup of tea and imagine thirty people walking into the room staring at you intently. Then try to imagine a million,” he said. Every month, his TikTok videos were getting forty million views. Kabvina’s videos, the algorithmic feed, and the rapacious audience formed a feedback loop. He called it “instant gratification”: “I can post on TikTok and in ten minutes I can check back and see thirty thousand people have watched it.”

卡布维纳的背景几乎不可能让他在社交媒体上出名。他出生于马拉维，他的父亲通过交换项目前往英国曼彻斯特郊外的一个小镇就读于一所工程学院。21 世纪初，卡布维纳六岁时，他的父亲带着他和母亲搬到了曼彻斯特。曼彻斯特这座城市无论从气温还是气候上来说都很冷。尽管他们搬到了一个马拉维人的小社区，但全家还是面临着来自英国当地人的种族歧视。“人们往我们家扔东西；人们会朝我吐口水。小时候，你对这些很不敏感，”他说。这种氛围加剧了一种疏离感。小时候，卡布维纳对学校并不热心，直到遇到了克拉克先生，一位严肃认真的数学老师，卡布维纳的成绩开始提高。有一天，数学班的另一个学生说他想成为一名会计，卡布维纳决定效仿他。与此同时，他的母亲继承了儿子强烈的完美主义倾向，她兼职做面包师。她让卡布维纳走上了一条高度独立的道路：他自己挑选食品杂货，自己准备午餐带去学校，自己熨衣服。他尤其喜欢烹饪，但他猜想父母永远不会同意他以此为职业道路。卡布维纳在大学学习数学和会计，还学习了电影研究——这些都为他未来的TikTok职业生涯提供了优势。但一份工作瑞银伦敦分行的职位申请因官僚主义而受阻，最终他回到了曼彻斯特。在那里，他开始从事调酒工作，并发现自己乐在其中。他从实习调酒师一路晋升，最终成为鸡尾酒比赛冠军和酒吧经理。

Little in Kabvina’s background would have suggested him for social media fame. He was born in Malawi, and his father attended an engineering college in the United Kingdom on an exchange program in a town outside of Manchester. When Kabvina was six years old, in the early 2000s, his father moved him and his mother to Manchester. The city was cold, both in temperature and temperament. Though they moved into a small Malawian community, the family faced racism from British locals. “People threw stuff at our house; people would spit on me. When you’re a kid you’re desensitized to it,” he said. The atmosphere contributed to a sense of alienation. As a child Kabvina was unenthusiastic in school until he encountered Mr. Clark, a math teacher whose no-nonsense style agreed with him, and Kabvina began improving on tests. One day, another student in the math class said he wanted to be an accountant, and Kabvina decided to follow his lead. At the same time, his mother, who had a strong perfectionist streak that her son inherited, was working part-time as a baker. She set Kabvina on a path of hyper-independence: he picked his own groceries, made his own packed lunch for school, and ironed his own clothes. He especially enjoyed cooking, but guessed that his parents would never approve of it as a career path. Kabvina studied math and accounting in college, and took up film studies as well—all strengths for his future TikTok career. But a job offer at UBS in London was derailed by bureaucracy, and he ended up back in Manchester, where he fell into bartending and found he enjoyed it. He progressed from trainee bartender to winner of cocktail competitions and bar manager.

在2010年代中期，社交媒体似乎并未影响Kabvina的生活，这在当时实属罕见。他不喜欢在餐厅里拍美食照片发到Instagram上，不过他喜欢拍些照片作为自己的私人档案。他的朋友们大多是年纪稍长的调酒师，对互联网不屑一顾，每次和他一起出去玩时，Kabvina掏出手机都会被他们斥责。他们说，我们活在当下；你抓住了它，你就毁了它。我那个年代，我们经常和人聊天。他的犹豫不决，部分源于英国人根深蒂固的羞耻感，以及不愿在公共场合出丑——而这恰恰是TikTok最推崇的行为。

Unusually for the mid-2010s, social media didn’t factor into Kabvina’s life. He wasn’t into Instagramming his food at restaurants, though he liked taking the photos for his own personal archive. His friends, mostly older bartenders, disdained the Internet and chided Kabvina whenever he pulled out his phone while they were hanging out. We’re in the moment; you’re capturing it, you’re ruining it, they said. Back in my day we used to talk to people. His hesitancy was partly due to the entrenched British sense of shame and a reluctance to look silly in public—which is coincidentally the act that the TikTok feed prizes most.

2020年的疫情封锁改变了一切。Kabvina的公寓位于曼彻斯特第一栋高层建筑的14层，这栋建筑的18层还设有游泳池、健身房和桑拿房，为他日新月异的摄影作品提供了一个完美的工作室。落地窗提供了理想的采光，无需使用电动聚光灯。最近翻新的简洁几何造型提供了一种单色照片背景，他的美食作品在这种背景下格外醒目。这间奢华的公寓让我想起了作家兼软件工程师Paul Ford在2014年写的一篇名为《美国房间》的文章。在文中，Ford描述了YouTube视频的典型背景，这种视频通常在郊区房屋里某个古怪的米色角落或地下室拍摄。“对我们大多数人来说，生活的背景都是灰白色墙壁交错的背景，”Ford写道。但作为其后继者的TikTok典型房间则更加华丽。墙壁仍然是白色的，但房间的装饰风格统一，并且由阳光而不是荧光灯照明。它符合 Instagram 不言而喻但无处不在的美学。

The pandemic lockdown of 2020 changed everything. Kabvina’s apartment, on the fourteenth floor of the first high-rise built in Manchester, which has a pool, gym, and sauna on the eighteenth floor, provided a perfect studio for his evolving shoots. The floor-to-ceiling windows made for ideal lighting, without the need for electric spotlights. And the clean geometry of a recent renovation provided a kind of monochrome photo backdrop against which his food presentations stood out. The generically luxurious apartment reminded me of an essay that the writer and software engineer Paul Ford wrote in 2014 called “The American Room.” In it, Ford describes the archetypal background of a YouTube video, the kind often shot in the odd beige corner or basement of some suburban house. “For most of us life happens against a backdrop of intersecting off-white walls,” Ford wrote. But the archetypal TikTok room, its successor, is fancier. The walls are still white, but the room is coherently decorated and lit by sunlight instead of fluorescents. It complies with the unspoken but ubiquitous aesthetics of Instagram.

Kabvina 在他的 TikTok 账号中构建了自己的叙事弧线，打造了一段社交媒体时代的英雄之旅。他研究了最受欢迎的账号。像 Charli D'Amelio 和 Emily Mariko 这样的网红，部分原因在于他们从默默无闻开始成名。“我注意到最大的趋势是……（粉丝们）希望有一个主角带领他们踏上这段旅程，”Kabvina 说。他还仔细研究了根据TikTok提供的数据，他优化了自己的烹饪视频。避免过多的对话或文字，让视频吸引了全球观众——他的美食无需翻译。（这是一个成功的策略；Mariko也因其无台词的烹饪视频而出名。）TikTok应用会向创作者显示用户在视频的哪个部分会跳到下一个视频。如果观众在第19秒就跳过了，Kabvina会回头检查表现不佳的部分，然后尝试在下一个视频中避免出现类似的问题。这些精准的数据让他能够时刻优化视频的参与度。

Kabvina built his own narrative arc into his TikTok account, creating a social-media-era hero’s journey. He studied the most popular accounts. Influencers like Charli D’Amelio and Emily Mariko became famous in part for getting famous, starting from anonymity. “The biggest trend I’d notice is…[followers] want a protagonist to take them on this journey,” Kabvina said. He also carefully optimized his cooking videos according to the data TikTok gave him. Avoiding too much speaking or text made them appealing to a global audience—his food needed no translation. (It was a successful strategy; Mariko also became famous for her speech-less cooking videos.) The TikTok app reveals to creators at which point in a video users tune out and flip to the next video. If viewers were skipping at nineteen seconds, Kabvina would go back and examine the underperforming section, and then try to avoid its problems in the next video. Such specific data allowed him to optimize for engagement at every moment.

Kabvina 喜欢这种细致的反馈和不断改进作品的迭代过程，这或许是他数学背景的传承。“我看到创作者对算法感到沮丧；他们认为算法出了问题，”他说。“责怪算法比试图说‘我的内容不够好’要容易得多。”对于独立创作者来说，算法取代了老板和绩效评估；它是一个实时的权威，衡量你是否成功适应了它对引人入胜内容的定义，而这个定义总是在变化。

Kabvina liked the granular feedback and the iterative process to improve his work, perhaps a holdover from his math background. “I see creators get frustrated with the algorithm; they’re assuming something’s wrong,” he said. “It’s a lot easier to blame the algorithm than to try to say, ‘My content isn’t that good.’ ” For independent creators, the algorithm takes the place of bosses and performance reviews; it’s a real-time authority gauging your success at adapting to its definition of compelling content, which is always shifting.

Kabvina 建立了一套自己的公式来评判自己的作品。他计算特定视频的观看次数，然后从中抽取 10%。这就是一部视频获得多少点赞才能算作成功：至少有十分之一的观众会被点赞所吸引。艺术家一直以来都以数字来衡量自己的作品，无论是广播剧的播出量、电影票的销量，还是博物馆展览的参观人数。但在 Filterworld 出现之前，创作品味——艺术家评价自己作品的方式，以及消费者评价作品欣赏的方式——从未如此受到数据和对注意力进行精细测量的影响。

Kabvina has established his own equation that he judges his work by. He takes the number of views a particular video gets, then takes 10 percent of it. That’s how many likes a video needs to qualify as successful: at least one-tenth of the viewers have to be compelled enough to take the action of hitting the Like button. Artists have always evaluated their work based on numbers, whether on-air radio plays, movie tickets sold, or museum exhibition attendance. But before Filterworld, creative taste—which is the way that artists evaluate their own work, as much as how consumers evaluate their appreciation of it—has never been so influenced by data and the granular measurement of attention.

环境文化

AMBIENT CULTURE

虽然奈杰尔·卡布维纳（Nigel Kabvina）在算法推送的压力下蓬勃发展，并不断利用算法推送来增加粉丝数量和获得赞助内容的机会，但同样的力量也作用于各种各样的创作者。正如算法推荐将用户按消费类别进行分类，取代了他们的个人品味一样，它们也对用户进行分类。文化产出被划分成不同的类别，而创作者则面临着挑战。这些类别构成了Filterworld的固有风格，其伟大之处在于优化，而非对未知领域充满戏剧性的创意飞跃。文化不断被数字平台生成的海量数据提炼，这些平台提供着受众参与内容、时间及方式的逐秒记录。预设的风格无处不在，却又令人麻木。同质化开始疏远消费者，而非取悦他们，而罪魁祸首正是“算法”。近年来，一种潜在的感觉浮现：算法文化肤浅、廉价，如同被多次复制的褪色影印件般堕落。这也是一种算法焦虑：当创造文化这种人类活动如此自动化时，真实性就变得不可能。

While Nigel Kabvina thrives on the pressure of an algorithmic feed and keeps surfing it to higher follower numbers and opportunities for sponsored content, the same forces are acting on all kinds of creators. Just as algorithmic recommendations slot users into categories of consumption, supplanting their personal tastes, they also sort cultural output into categories, which creators run up against. These categories are the house styles of Filterworld, in which greatness is defined by optimization rather than dramatic creative leaps into the unknown. Culture is continuously refined according to the excesses of data generated by digital platforms, which offer a second-by-second record of what audiences are engaging with and when and how. The preset styles are omnipresent and stultifying. Homogenization is beginning to alienate consumers rather than entertain them, and the entity to blame is “the algorithm.” In recent years, an underlying sense has emerged that algorithmic culture is shallow, cheap, and degraded in the washed-out manner of a photocopy copied many times over. This, too, is a form of algorithmic anxiety: the feeling that, when such a human endeavor as making culture is so automated, authenticity becomes impossible.

关于这种普遍存在的肤浅现象的抱怨越来越频繁，也越来越强烈。我开始收集这些抱怨，以记录这种日益增长的反感。诗人艾琳·迈尔斯（Eileen Myles）曾说，创作过程与数字技术密不可分：“你可能不使用社交媒体，但它在利用你。无论你喜不喜欢，你都在发推文。” 剧作家兼小说家阿亚德·阿赫塔尔（Ayad Akhtar）将我们的“点击诱饵意识”描述为一种被训练成与信息流中任何旨在引发共鸣的内容互动的意识。“对算法的崇拜正在摧毁创意产业，”电视编剧科德·杰斐逊（Cord Jefferson）抱怨道。“文化不再是被创造出来的。它只是从现有文化中提炼出来，经过提炼，然后被重新灌输给我们。算法切断了新发现的可能性，”反科技生活方式影响者保罗·斯卡拉斯（Paul Skallas）写道，他哀叹2010年代电影续集和漫威超级英雄系列电影的无休止续集泛滥。斯卡拉斯将这种缺乏创新的文化称为“停滞不前的文化”。 “算法将未来限制在过去，”未来学家贾罗德·拉尼尔（Jarod Lanier）说道。正如已故英国哲学家马克·费舍尔（Mark Fisher）所写：“二十一世纪被一种压倒性的有限感和疲惫感所压迫。它感觉不像未来。”

Complaints regarding this pervasive shallowness have popped up increasingly often and with increasing intensity. I’ve taken to collecting them to document the growing antipathy. The poet Eileen Myles said that it is impossible to separate the creative process from digital technology: “You may not use social media, but it’s using you. You’re writing in tweets, like it or not.” The playwright and novelist Ayad Akhtar described our “click-bait consciousness,” trained to interact with anything in the feed designed to be triggering. “The worship of algorithms is mutilating creative industries,” the television writer Cord Jefferson complained. “Culture is no longer made. It is simply curated from existing culture, refined, and regurgitated back at us. The algorithms cut off the possibility of new discovery,” wrote Paul Skallas, an anti-technology lifestyle influencer, bemoaning the 2010s’ plague of movie sequels and endless continuations of Marvel superhero franchises. Skallas labeled this lack of innovation “stuck culture.” “Algorithms are limiting the future to the past,” the futurist Jarod Lanier said. As the late British philosopher Mark Fisher wrote, “The twenty-first century is oppressed by a crushing sense of finitude and exhaustion. It doesn’t feel like the future.”

这种认为文化停滞不前、受困于千篇一律的观念，确实是由于算法推送的无处不在。但这并不意味着创新没有发生；而是创新只会朝着推送的方向发展，鼓励发展。就像奈杰尔·卡布维纳（Nigel Kabvina）的烹饪视频一样，这些产品服务于数字平台的结构。毕竟，算法文化的完美之作几乎是故意设计的，毫无趣味。它的空虚有时是字面意义上的。2014年，一支名为Vulfpeck的乐队在Spotify上上传了一张包含十首无声曲目的专辑，并将其命名为《Sleepify 》。这是一首基于约翰·凯奇（John Cage）1952年作品《4'33"》的数字即兴重复乐段，每首曲目的标题都以一个个递增的“Z”开头。乐队利用推特（Twitter）等其他数字平台，鼓励听众在睡觉时重复播放这首歌，从而从Spotify获得流媒体版税，而Spotify无法判断用户是否真正在听音乐。这张催眠专辑至少获得了两万美元的流媒体版税，但Spotify要求乐队将其下架，理由是它“违反了他们的内容条款”。 （作为回应，乐队录制了一首名为“官方声明”的曲目并上传了它。）尽管它最终消失了，就像所有对抗数字平台的内容一样，尽管它是一个字面上的空白，但Sleepify还是获得了某种成功。

This perception that culture is stuck and plagued by sameness is indeed due to the omnipresence of algorithmic feeds. But it’s not that innovation isn’t happening; it’s that innovation is improving only in the direction of the feed, encouraging the development of products that serve the structure of digital platforms, as Nigel Kabvina’s cooking videos do. The perfect piece of algorithmic culture, after all, is almost intentionally uninteresting. Its emptiness can at times be literal. In 2014, a band called Vulfpeck uploaded an album of ten tracks of silence on Spotify and titled it Sleepify. A digital riff on John Cage’s 1952 composition 4'33", each track was titled with an increasing number of Z’s. Using other digital platforms like Twitter, the band encouraged their listeners to play it on repeat while they slept, earning the band streaming royalties from Spotify, which had no way of telling whether a user was truly paying attention to the music or not. The soporific album earned at least twenty thousand dollars in streaming royalties, but Spotify requested that the band take it down, because it “violated their terms of content.” (In response, the band recorded a track called “Official Statement” and uploaded that, too.) Though it eventually disappeared, like all content that goes up against digital platforms, and despite being a literal void, Sleepify found a kind of success.

尽管“氛围音乐”出现于20世纪，但与21世纪文化息息相关的概念之一，却是作曲家布莱恩·伊诺（Brian Eno）创造的“氛围音乐”。伊诺在1978年发行的专辑《机场音乐》（Music for Airports）的封套说明中，为这一新流派命名。他写道：“氛围音乐必须既引人入胜又易于忽略。氛围被定义为一种氛围，或一种周围环境的影响：一种色调。” 伊诺的音乐有意营造氛围感。《机场音乐》收录了一组缓慢柔和的合成器乐曲，它们如同海浪拍打沙滩时缓缓退去般起伏，这种氛围之声与机场这种转瞬即逝、空灵的空间格格不入。聆听这张专辑时，它会轻柔地为你的感官环境增添色彩，但又不会造成任何干扰——你可以一边听音乐一边工作，或者进行对话，或者将其视为一件艺术品进行冥想。它能够满足各种形式的关注，并适应各种目的。在“过滤世界”（Filter world），文化正变得越来越具有氛围感。就像《Sleepify》一样，它的设计初衷就是让人忽略；或者，就像漫威系列电影一样，它的任何一个瞬间或片段都毫无意义，因为总有更多精彩等待我们去品味。当我们拥抱氛围时，我们便失去了有限和离散的意义。

Though it emerged in the twentieth century, one of the most relevant concepts for culture in the twenty-first century is the composer Brian Eno’s coinage “ambient music.” Eno named the new genre in the liner notes of his album Music for Airports in 1978. Ambient music “must be as ignorable as it is interesting,” he wrote. “An ambience is defined as an atmosphere, or a surrounding influence: a tint.” Eno made music that was intentionally ambient. Music for Airports was a collection of slow, soft synthesizer compositions that flowed in and out like waves slowing ebbing as they hit beach sand, a fitting atmospheric sound for a transient, ethereal space like an airport. When you listen to the album, it lightly colors your sensory environment without getting in the way—you can do work while listening to it, or carry on a conversation, or meditate on it as an artwork. It rewards all forms of attention, adapting to any purpose. In Filterworld, culture is becoming more ambient. Like Sleepify, it’s designed to be ignored, or, like the Marvel movie franchise, no single moment or fragment of it is particularly significant because there is always more to be consumed. When we embrace ambience, we lose the meaning of the finite and the discrete.

TikTok 的“为你推荐”动态就是这种趋势的一个很好的例子。用户无需过于关注任何内容，信息流就会自动加载，并根据用户之前的操作进行个性化定制。这些信息流不会疏远用户的个人品味，也不会对用户产生强烈的吸引力。这样的信息流提升了氛围的质量，因为它们能够维持持续的互动基线；用户永远不会停止信息流。然而，如果像在TikTok上狂刷一样长时间地浏览这些信息，就会让人产生一种非人性化的感觉：你正在成为算法信息流所认为的那个你吗？或者你本来就是那样的你吗？

The TikTok “For You” feed is a good example of this tendency. The feed flows by without the user needing to pay attention to any one thing too closely, since another is always about to load, personalized to some degree by their previous actions. There is nothing to alienate personal taste but also nothing to deeply compel it. Such feeds encourage the quality of ambience because they allow for a consistent baseline of engagement; the user never stops the flow. Yet consuming it for too long at a stretch, like during a TikTok binge, leads to a feeling of depersonalization: Are you becoming the person the algorithmic feeds perceive you as, or were you already?

继布莱恩·伊诺的“机场音乐”之后，YouTube 频道“低保真嘻哈电台——放松/学习的节奏”正式上线。该频道由一位名叫 Dmitri 的 DJ（用户名为 ChilledCow）于 2015 年创建。它是一个全天候 24/7 不间断的直播频道，播放中速、带有原声电子乐风格的音乐，听起来时而朦胧，时而怀旧，仿佛未来的音乐从一台静电干扰的收音机里传出来。该频道单次播放时长超过两万小时，拥有超过一千两百万订阅者；它或许是我们这个时代最受欢迎的音乐平台之一。然而，这些音乐却令人难以忘怀：歌曲之间几乎没有区别，音乐低调的一致性使其正如标题所示，成为放松、学习——或者说，睡觉——的完美选择。 （受吉卜力工作室电影启发，动画中一个戴着大耳机坐在书桌前学习的女孩提供了合适的视觉氛围。）正如 Eno 所说，它可以被忽略，也可以被积极倾听——但它大多被忽略了。

The successor to Brian Eno’s Music for Airports is the YouTube stream “lofi hip hop radio—beats to relax/study to,” which was created by a DJ named Dmitri going by the username ChilledCow in 2015. It’s a 24/7 stream, like a live radio channel, of mid-tempo, acoustic-inflected electronica that often sounds hazy or nostalgic, like music from the future coming through a staticky radio. The channel has played for stretches of over twenty thousand hours at a time and has over twelve million subscribers; it might be one of the most popular music outlets of our era. And yet the music is utterly forgettable: the songs are almost indistinguishable from one another, and the music’s unobtrusive consistency makes it perfect, as the title suggests, for relaxing or studying—or literally sleeping. (An anime-style animation of a girl wearing large headphones studying at her desk, inspired by a Studio Ghibli movie, provides the appropriate visual mood.) As Eno suggested, it can be ignored or actively listened to—but it mostly gets ignored.

像奈杰尔·卡布维纳 (Nigel Kabvina) 早期那些充满氛围的片段和无声的烹饪技艺这样的视频，传达了一种没有具体含义的情绪，而这完全取决于消费者的解读。它们可以有任何含义，也可以没有任何含义。许多流媒体电视节目也注重氛围营造，淡化叙事，转而强调引人入胜的氛围，默契地为观众留下了观看手机的空间，让他们不会错过太多精彩内容。剧情喜剧《艾米丽在巴黎》于 2020 年疫情隔离期间在 Netflix 上首播，它或许是巴黎的屏幕保护程序，庆祝女主角在社交媒体上发布她在巴黎精彩生活的片段。当我们不看流媒体电视或沉迷于社交网络应用时，通过……AirPod 耳机提供环境喋喋不休的声音，这种噪音代表着没有现实生活中的社交接触的社交刺激。

Videos like Nigel Kabvina’s early atmospheric clips and speech-less cooking feats impart a mood without concrete meaning, which is left up to the consumer’s interpretation. They can mean, or not mean, anything. Many streaming television shows are ambient, too, deemphasizing narrative in favor of compelling atmosphere, tacitly leaving room for the viewer to look at their phone while watching without missing too much. Debuting on Netflix in 2020 during pandemic quarantine, the dramedy series Emily in Paris may as well have been a screen saver of Paris, celebrating its heroine’s accomplishments of posting scenes from her fabulous life in the city on social media. When we’re not watching streaming television or hypnotized by social-networking apps, podcasts piped in through AirPod headphones provide ambient chatter, noise that represents social stimulus without real-life social contact.

很多情况下，通过算法推送传播的文化要么是为了制造感官空虚，要么就是为了将其扁平化为生活的背景，将艺术的地位不知不觉地降格为墙纸。虽然个性化推送会引发算法焦虑，但它们提供的唯一慰藉就是这种氛围文化，这种文化可以让人感觉个性化，但实际上并非如此——如果你不必因为消费的艺术太平淡而过多思考，你就不需要担心它是否真正代表你。就像一种企业化的佛教，对焦虑的隐含答案是学会首先不要渴望差异化，而要满足于摆在你面前的一切。品味的培养不被鼓励，因为品味在最大化参与度方面效率低下。

In so many cases, the culture disseminated through algorithmic feeds is either designed to produce a sensory void or to be flattened into the background of life, an insidious degradation of the status of art into something more like wallpaper. While personalized feeds create algorithmic anxiety, the only salve they provide is this form of ambient culture, which can feel personal without actually being so—if you don’t have to think too much about the art you’re consuming because it’s so bland, you don’t need to worry about if it truly represents you or not. Like a corporatized form of Buddhism, the implied answer to anxiety is to learn not to desire differentiation in the first place, to simply be satisfied with whatever is placed in front of you. The cultivation of taste is discouraged because taste is inefficient in terms of maximizing engagement.

虽然品味的阻碍和腐败的个性化策略可能感觉像是个人问题——用户必须更加努力才能确定自己真正喜欢什么——但它也迅速升级为大规模的社会问题。当数百万消费者被巧妙地误导，最终被灌输他们所消费的内容时，某些文化就被扼杀在曝光和融资的边界之外。这是因为资本的流向发生了变化，更容易流向那些适应算法推送的事物。这种影响在选择观看的电视节目或购买的衣服方面显而易见，但它也影响着我们生活中更宏观的行为，比如我们去的餐馆、我们旅行的地方，以及我们如何与邻里和社区成员互动。

While the obstruction of taste and the tactic of corrupt personalization may feel like individual problems—users must work harder to identify what they truly like—it also quickly scales up into massive social issues. When millions of consumers are subtly misled and thus ultimately fed what they consume, certain kinds of culture are choked off from exposure and financing. That is because the flow of capital changes and moves more easily to things that adapt to algorithmic feeds. The effect is blatant in the context of selecting TV shows to watch or clothes to buy, but it also influences larger-scale actions in our lives, like the restaurants we go to, the places we travel, and how we interact with our neighborhoods and community members.

寻找普通的咖啡店

SEEKING THE GENERIC COFFEE SHOP

过滤世界并不局限于我们屏幕上的数字体验，它更是一股塑造物理世界的无所不在的力量。由于算法系统影响着我们作为个体所消费的文化类型，塑造着我们的个人品味，它们也影响着我们倾向于选择哪些地方和空间。而我们的偏好所指向的地方，那些希望向我们销售商品或吸引我们注意力的企业也会随之而来，迎合这些偏好。正如Netflix、Spotify和Instagram通过算法优先推荐符合各自平台结构的特定数字内容一样，其他应用程序也会将我们的注意力引导到同样符合平台激励机制的地方。Airbnb会引导用户找到符合其需求的算法所反映的房屋租赁信息；谷歌地图通过在数字地图上突出显示一组个性化的本地机构来突出显示它们；Yelp和Foursquare则会将用户评论和参与度汇总到餐厅、酒吧和咖啡馆的排名列表中。想象“信息流”存在于屏幕之外似乎有些奇怪，但这些应用程序的工作原理就像Netflix算法在物理空间中的主页一样。您可以向下滚动列表，选择您想要的体验。这些应用让我们的现实生活体验与数字世界一样顺畅。

Filterworld is not limited to digital experiences on our screens. It is a pervasive force that shapes the physical world, too. Because algorithmic systems influence the kinds of culture we consume as individuals, molding our personal tastes, they also influence what kinds of places and spaces we gravitate toward. And where our preferences go, businesses hoping to sell us things or grab our attention will follow, catering to those preferences. The same way that Netflix, Spotify, and Instagram algorithmically prioritize certain kinds of digital content that fit with each platform’s structure, other apps direct our attention toward places that similarly meet the platforms’ incentives. Airbnb guides users toward home rentals that match the algorithmic reflection of their desires; Google Maps highlights a personalized set of local institutions by emphasizing them on the digital map; and Yelp and Foursquare collate user reviews and engagement into ranked lists of restaurants, bars, and cafés. It’s strange to think of a “feed” existing outside of a screen, but these apps work like an algorithmic Netflix home page for physical space. You can scroll down the list and select the experience you want to have. These apps make our IRL experiences just as frictionless as our digital ones.

在2010年代的大部分时间里，我都是Yelp的忠实用户，这是一款用于查找和评价餐厅和其他本地商家的应用程序。红白相间的界面成为了我信赖的信息来源。推荐。住在布鲁克林的时候，我每隔一周就会打开它，看看公寓附近有没有新开的咖啡馆，或者看看那些我还没去过的地方的评价。在报道旅行中，需要找个地方工作或在会议间隙打发时间时，我也会用这个应用。在柏林、京都和雷克雅未克，我搜索咖啡店，然后快速浏览 Yelp 的列表，这个列表是按咖啡馆的星级排序的——星级反映了该应用其他用户对每个地方的喜爱程度。

For most of the 2010s, I was a religious user of Yelp, an app for finding and reviewing restaurants and other local businesses. The red-and-white interface became a trusted source of recommendations. When I lived in Brooklyn, I would open it every other week to see if new cafés had opened near my apartment or to check in on reviews of places I hadn’t been yet. I also turned to the app when I was on reporting trips and needed a place to work or pass the time between meetings. In Berlin, Kyoto, and Reykjavík, I searched for coffee shops, and quickly scrolled through Yelp’s list, which was sorted by the cafés’ star rating—a reflection of how much the app’s other users had liked each spot.

我经常在搜索栏里输入“潮人咖啡店”作为缩写，因为 Yelp 的搜索算法总能准确地理解我的意思。像我这样二十多岁（当时），西方人，互联网思维的千禧一代，敏锐地意识到自己的品味，才会想去那种咖啡馆。我总能在搜索结果中快速找到一家具备以下条件的咖啡馆：店面窗户宽敞明亮，采光充足；工业尺寸的木桌方便就座；明亮的室内空间，墙壁刷成白色或铺着地铁瓷砖；还有无线网络，方便写作或打发时间。当然，咖啡本身也很重要，在这些咖啡馆，你可以放心地喝上一杯用时尚的轻度烘焙浓缩咖啡（比传统的深度烘焙具有更浓郁的果味）制作的卡布奇诺，选择牛奶的种类（包括全脂、豆奶、杏仁奶、火麻仁奶和燕麦奶，种类越来越多），还可以制作精美的拿铁艺术（在咖啡顶部倒入玫瑰花结图案和蒸牛奶，变成了一种潮人的品牌标志）。最用心的咖啡馆会提供“馥芮白”（一种澳大利亚卡布奇诺变体）和牛油果吐司，这是一道简单的菜肴，也源自澳大利亚，在 2010 年代成为千禧一代消费者偏好的提喻。臭名昭著的头条新闻将千禧一代对昂贵牛油果吐司的偏爱归咎于他们无力在高档化城市购买房地产。

I often typed “hipster coffee shop” into the search bar as a shorthand because Yelp’s search algorithm always knew exactly what I meant by the phrase. It was the kind of café that someone like me, a Western, twentysomething (at the time), Internet-brained millennial acutely conscious of their own taste, would want to go to. Inevitably, I could quickly identify a café among the search results that had the requisite qualities: plentiful daylight through large storefront windows; industrial-size wood tables for accessible seating; a bright interior with walls painted white or covered in subway tiles; and Wi-Fi available for writing or procrastinating. Of course, the actual coffee mattered, too, and at these cafés you could be assured of getting a cappuccino made from fashionably light-roast espresso (which has fruitier flavor notes than traditional dark roast), your choice of milk variety (including whole, soy, almond, hemp, and oat, in a proliferating list), and elaborate latte art (a rosette pattern poured into the top of the coffee with steamed milk, which turned into a kind of hipster brand logo). The most committed among the cafés would offer a “flat white” (an Australian cappuccino variant) and avocado toast, a simple dish, also with Australian origins, that over the 2010s became synecdochic for millennial consumer preferences. Infamous headlines blamed millennials’ predilection for expensive avocado toast for their inability to buy real estate in gentrifying cities.

这些咖啡馆都采用了类似的美学风格，提供类似的菜单，但它们并非像星巴克那样被母公司强迫这样做。对于一家公司而言，严格的统一性确保了其各门店的效率、熟悉度和可靠性，从而提升了顾客忠诚度和盈利能力。相反，尽管它们地理位置分散，总咖啡馆彼此独立，都朝着同一个终点漂移。这种千篇一律的景象令人震惊，也令人耳目一新，让人觉得乏味。我感到困惑和错乱，就像坐夜班飞机降落在另一个国家，感觉有点不真实一样。这似乎太容易了。

These cafés had all adopted similar aesthetics and offered similar menus, but they hadn’t been forced to do so by a corporate parent, the way a chain like Starbucks replicated itself. For a corporation, strict uniformity ensures efficiency, familiarity, and dependability across its locations—driving customer loyalty and profitability. Instead, despite their vast geographical separation and total independence from each other, the cafés had all drifted toward the same end points. The sheer expanse of sameness was too shocking and new to be boring. I felt bemused and dislocated by it, the way it feels slightly unreal to take an overnight flight and land in a different country. It seems too easy.

当然，这种文化全球化的例子可以追溯到有记载的文明时期，从遍布古罗马帝国的大理石神庙和澡堂的同质性，到各种无处不在的殖民主义和全球移民的象征：18世纪随处可见的英式奶茶，千篇一律的爱尔兰酒吧，以及移民开设的中餐馆。事实上，法国社会学家加布里埃尔·塔尔德早在1890年就曾抱怨过同质性，因为随着客运列车的兴起，旅游业在欧洲各地兴起。正如他在《模仿法则》中所写：

Of course, there have been examples of such cultural globalization going back as far as recorded civilization, from the homogeneity of marble temples and bathhouses dotted across ancient Rome’s empire to various omnipresent symbols of colonialism and global migration: the eighteenth-century ubiquity of the milky British cup of tea, the sameness of the Irish pub or the Chinese restaurant opened by immigrants. In fact, the French sociologist Gabriel Tarde complained about homogeneity in 1890 because of the travel industry emerging across Europe with the rise of passenger trains. As he wrote in The Laws of Imitation:

现代大陆游客会发现，特别是在大城市和上层阶级中，酒店餐饮和服务、家用家具、服装和珠宝、戏剧公告以及商店橱窗中的商品都千篇一律。

The modern continental tourist will find, particularly in large cities and among the upper classes, a persistent sameness in hotel fare and service, in household furniture, in clothes and jewelry, in theatrical notices, and in the volumes in shop windows.

相互联系的地方，仅仅因为产品、人员和思想的流动，在某些方面逐渐变得相似。交流越快，相似性就越快形成。但21世纪那些千篇一律的咖啡馆，其独特之处不仅在于其细节的匹配，还在于它们仿佛是自然而然地从其所在位置衍生而来。它们是引以为豪的本地特色，常常被描述为“地道”，而我也曾过度使用过这个形容词。旅行时，我总是想找个“地道”的地方喝点东西或吃顿饭。然而，如果这些地方都如此相似，那么它们究竟是向着什么而“地道”呢？“地道”通常意味着与某个根源的直接联系，一个历史和意义的稳定源泉。我最终得出的结论是，它们都与新的数字地理网络真实地连接在一起，彼此连接。通过社交网络实时推送。它们对于互联网来说是真实的，尤其是2010年代算法推送的互联网。

Places that connect gradually grow to resemble each other in certain ways simply because of their interconnection, in the sense of the movement of products, people, and ideas. The faster the exchange, the faster the similarity sets in. But the twenty-first-century generic cafés were remarkable in the specificity of their matching details as well as the sense that each had emerged organically from its location. They were proud local efforts that were often described as “authentic,” an adjective that I was also guilty of overusing. When traveling, I always wanted to find somewhere “authentic” to have a drink or eat a meal. If these places were all so similar, though, what were they authentic to, exactly? Authenticity usually implies a direct connection to a certain root, a stable source for history and meaning. What I eventually concluded was that they were all authentically connected to the new network of digital geography, wired together in real time by social networks. They were authentic to the Internet, particularly the 2010s Internet of algorithmic feeds.

2016年，我为The Verge撰写了一篇题为《欢迎来到空域》（Welcome to AirSpace）的文章，描述了我对这种千篇一律现象的最初印象。“空域”是我创造的词，指的是数字平台创造的奇特的无摩擦地理环境，在这种环境中，你可以在不同地点之间移动，而无需超越应用程序的界限。这个词部分是对Airbnb的致敬，因为它能够实现全球旅行，但也受到了这些地方给我的那种虚无缥缈和不真实感的启发。它们似乎与地理位置如此脱节，以至于它们可以飘走并降落到任何其他地方。当你身处其中时，你可能身处任何地方。

In 2016, I wrote an essay titled “Welcome to AirSpace” for the Verge, describing my first impressions of this phenomenon of sameness. “AirSpace” was my coinage for the strangely frictionless geography created by digital platforms, in which you could move between places without straying beyond the boundaries of an app. The word was partly a riff on Airbnb, with its ability to enable global travel, but it was also inspired by the sense of vaporousness and unreality that these places gave me. They seemed so disconnected from geography that they could float away and land anywhere else. When you were in one, you could be anywhere.

我的理论是，所有通过应用程序互联的实体场所——相当于塔德的泛欧列车的当代版本——都有一种彼此相似的方式。就咖啡馆而言，Instagram 的兴起让世界各地的咖啡馆老板和咖啡师能够实时关注彼此，并通过算法推荐逐渐开始消费同类的咖啡馆相关内容。一家咖啡馆老板的个人品味也会逐渐向其他人的喜好靠拢，最终汇聚成一个净平均值。在商业等式的另一端，Yelp、Foursquare 和谷歌地图将像我这样的顾客——他们也能在 Instagram 上追随流行的审美——吸引到符合他们心意的咖啡馆，方法是将这些咖啡馆放在搜索结果顶部或在地图上突出显示。为了吸引互联网塑造的庞大客户群体，更多的咖啡馆需要采用这些平台上已经占据主导地位的审美。适应常态不仅仅是追随时尚，而是一项商业决策，消费者会对此表示赞赏。当一家咖啡馆的视觉效果足够赏心悦目时，顾客会乐于在自己的Instagram上发布它，以此来炫耀自己的生活方式，这不仅提供了免费的社交媒体广告，还吸引了新顾客。就这样，审美优化和同质化的循环持续进行。

My theory was that all the physical places interconnected by apps—the contemporary equivalent of Tarde’s pan-European trains—had a way of resembling one another. In the case of the cafés, the growth of Instagram gave international café owners and baristas a way to follow each other in real time and gradually, via algorithmic recommendations, begin consuming the same kinds of café-related content. One café owner’s personal taste would drift toward what the rest of them liked, too, eventually coalescing into a net average. On the other side of the business equation, Yelp, Foursquare, and Google Maps drove customers like me—who could also follow the popular aesthetics on Instagram—toward cafés that conformed with what they wanted to see by putting those cafés at the top of searches or highlighting them on a map. To court the large demographic of customers molded by the Internet, more cafés needed to adopt the aesthetics that already dominated on the platforms. Adapting to the norm wasn’t just following fashion but making a business decision, one that the consumers rewarded. When a café was visually pleasing enough, customers felt encouraged to post it on their own Instagram in turn as a lifestyle brag, which provided free social media advertising and attracted new customers. Thus the cycle of aesthetic optimization and homogenization continued.

我最近在罗德岛州纽波特体验到了这种整体效果，纽波特是一座历史悠久的海滨度假小镇，部分原因是海盗活动，当地大多数设计都带有复古航海风格人工制品。氛围就像一艘格外奢华的海盗船。当我寻找避暑之地时，一家名为 Nitro Bar 的咖啡馆在我的谷歌地图上以一个大圆点格外醒目，因为算法预测我会喜欢它。它并非航海风格，而是拥有完美的“空域”美学：吊灯、悬浮式木质搁板，大理石台面上还有镀铜龙头系统，供应最新款的氮气冷萃饮品。谷歌高亮显示的照片让我预览了店内装潢，我还翻看了它的 Instagram 账号，该账号拥有超过两万名粉丝。这下我终于放心了，走进了实体世界的咖啡馆，感激地点了一杯卡布奇诺，拿铁艺术堪称完美。与其说是我偶然发现了这家咖啡馆，不如说是它找到了我。算法推荐根据我之前的数据，大致估算了我的口味，然后自动化处理，最后反馈给我。它们为我提供了一条通往 Nitro Bar 的近路。市中心街道上其他那些更俗气、更简陋、或许更有历史感的咖啡馆，我完全可以无视。

I recently experienced this overall effect in Newport, Rhode Island, an historic seaside resort town, founded in part on piracy, where most of the local design is styled with vintage nautical artifacts. The vibe is like a particularly luxurious pirate ship. When I looked for a spot to take a break from the heat, a café called the Nitro Bar was given particularly prominent placement with a big dot on my Google Maps app because I was algorithmically predicted to like it. Instead of maritime, it had the perfect AirSpace aesthetic, with pendant lamps, floating wooden shelves, and a copper-plated tap system on the marble countertop, which dispensed the latest in nitro cold brew drinks. Photos highlighted by Google gave me a preview of what the interior looked like, and I also flipped to its Instagram account, which had over twenty thousand followers. Thus reassured of its generic quality, I walked over to the café in the physical world and gratefully ordered a cappuccino, which came with perfect latte art. It felt less like I had stumbled across the café than it had found me. The algorithmic recommendations had approximated my taste based on my previous data, automated it, and then served it back to me. They provided a physical shortcut, rerouting my path toward the Nitro Bar. The rest of the cafés on the downtown street—kitschier, less polished, perhaps more historic—I could safely ignore.

当我找到一家这样的咖啡馆时，无论是想快速喝点咖啡，还是想长时间工作，我都感到很舒服，甚至有点自在。我确信自己能在那里写出最好的作品，因为这里没有干扰，我可以安心地拥有我喜欢的一切，而这在一家不那么普通、设计更混乱的咖啡馆里可能就无法实现。这也是一种自命不凡：我觉得自己融入了这些“边陲之地”。它们反映了我的品味和志向，因为我经常旅行，并以在世界各地工作为荣。但认同一个空洞的符号也让我感到很奇怪。

When I found one of these coffee shops, whether to grab a quick dose of caffeine or work for a prolonged period, I felt comforted and paradoxically at home. I was certain that I could do my best writing there because I was free of distractions and I could rely on having everything I liked, which may not have been true of a less generic, more chaotically designed café. It was also a form of pretension: I felt that I fit in in these outposts. They reflected my tastes and aspirations, as someone who traveled a lot and took a cosmopolitan pride in working everywhere. But identifying with a literally empty symbol also struck me as weird.

从某种程度上来说，完美的普通咖啡馆就像是新Word文档或网站背景的空白处：你投射的内容填满了它。有时，寻找它们感觉就像一场朝圣之旅，就像一位建筑游客寻找哥特式大教堂。当一家咖啡馆将普通美学表现得特别好，或者增添了一些新意时，我会欣喜若狂，享受着这种均衡的空白。帕蒂·史密斯写道，随着年龄的增长，她仍然会去咖啡馆写作，每次都会点一杯卡布奇诺，但并不需要真的喝掉卡布奇诺；它就摆在她笔记本旁边的桌子上，足以激发我的创造力。对我来说，这个空间本身就足以激发我的灵感。

In a way, the perfect generic café was like the blank space of a fresh Word document or a website background: what filled it was what you projected upon it. At times finding them felt like a pilgrimage, similar to an architecture tourist seeking out Gothic cathedrals. When a café executed the generic aesthetic particularly well, or innovated with some added novelty, I was giddily pleased and relished the well-balanced blankness. Patti Smith has written that as she got older, she still went to cafés to write and ordered a cappuccino each time, but didn’t need to actually drink the cappuccino; its presence on the table next to her notebook was enough to spark creativity. For me, the space itself was inspiration enough.

咖啡馆为我的数字地理学理论提供了一个完美的测试案例。它们是消费空间，特定人群中也非常活跃于互联网的人通过消费来表达他们的个人愿望。咖啡馆融合了建筑、室内设计和餐具等美学理念。它们展示了饮料和食品的流行趋势。它们播放着精选的柔和氛围音乐，例如低保真节拍。每家咖啡馆都汇聚了各种当代品味——用瓦格纳的话来说，它们堪称艺术品，指的是一种完全沉浸式的、调动所有感官的美学创作。咖啡馆成了我用来检验互联网对文化品味和消费模式影响的“煤矿中的金丝雀”。它们正是互联网影响最为明显的地方。

Coffee shops provided a perfect test case for my theory of digital geography. They were spaces of consumption, in which members of a certain demographic, who were also very active on the Internet, expressed their personal aspirations by spending money. The café space integrated aesthetic decisions across architecture, interior design, and tableware. They showcased trends in both beverages and food. They included a particular selection of music with soft ambient soundtracks, like lo-fi beats. Each café comprised a temple to all forms of contemporary taste—they were total works of art, to use Wagner’s term for a fully immersive aesthetic creation that engages every sense. Cafés turned out to be my canary in the coal mine for examining the impact of the Internet on cultural tastes and consumption patterns. They were where its influence was felt most blatantly.

正如17世纪早期西方的咖啡馆为民主和平等思想的传播提供了平台，将不同阶层的人聚集在一个实体空间里，2010年代的咖啡馆也创造了一种社会组织形式。它们为日益壮大的“零工”和数字创意工作者提供了一个休闲的聚会场所，这些人的日程安排与传统办公室朝九晚五的作息时间格格不入，也完全缺乏办公室的基础设施。（我就是其中之一，为了撰写计件文章，我辗转于布鲁克林的各个咖啡馆之间，结识其他常客。）除了通过数字平台建立联系之外，这个地理区域还服务于那些在数字平台上找到工作的人，无论是Uber的司机，还是Fiverr上的自由平面设计师。

Just as early Western coffee shops in the seventeenth century provided a venue for the proliferation of democratic and egalitarian ideas, by mingling different classes in one physical space, the cafés of the 2010s also created a form of social organization. They offered a casual gathering point for the growing economy of “gig workers” and digital creative laborers, people whose schedules didn’t align with the usual nine-to-five of the traditional office, and who lacked the infrastructure of an office entirely. (I was very much one of those people, drifting between various Brooklyn cafés to write piecework articles, getting to know the faces of other regulars.) In addition to being connected by digital platforms, this geography was also serving those who found work on digital platforms, whether drivers on Uber or freelance graphic designers on Fiverr.

2016年，我的“空旷空间”文章发表后，读者们纷纷接受了我创造的这个术语，并在自己的日常生活中观察到它。他们通过电子邮件向我发送了一些“空旷空间”风格的咖啡馆案例，并惊叹于这种风格如此普遍。虽然这种风格在咖啡馆尤为明显，但在共享办公空间、初创企业办公室、酒店和餐厅中也能找到同样的感觉——所有这些空间都让人暂时消磨时间，炫耀美感，将物理空间转化为一种产品。

When my AirSpace essay was published in 2016, readers embraced the term I had coined, observing it in their own day-to-day experiences. They emailed me examples of cafés that were “AirSpacey” and marveled at how ubiquitous the style was. Though it was particularly identifiable in cafés, the same sensibility could be found in coworking spaces, start-up offices, hotels, and restaurants—all spaces where time was temporarily spent and aesthetic was flaunted, where physical space was turned into a product.

然而，随着岁月的流逝，我意识到“空域”与其说是一种特定的风格，不如说是一种我们生存的状态，超越了单一的审美潮流。如同所有时尚一样，2010年代中期那种视觉风格也逐渐衰落。白色的地铁瓷砖开始显得过于老套，就像我童年时期家里的复合地板台面给人的感觉就像九十年代的风格一样，它们被色彩鲜艳或纹理更丰富的瓷砖所取代。早期布鲁克林高地伐木工人粗犷的风格，以及改造后的工业家具，让位于精心设计的斯堪的纳维亚式的中世纪现代主义，细腿椅子和细木工制品成为主流。在2010年代后期，主流审美变得更冷淡、更简约，水泥台面和粗糙的几何盒子取代了椅子。像用生锈的管道装置制成的灯具之类的装饰被抛弃，取而代之的是室内植物（尤其是多肉植物）和质感丰富的纤维艺术，这更让人联想到西海岸的波西米亚风情，而非贫瘠的纽约市。这种与布鲁克林的联系逐渐淡化（2020年疫情过后，布鲁克林本身被认为不如曼哈顿下城那么受欢迎），这种千篇一律的风格与其说与某个地方有关，不如说与Instagram和TikTok等数字平台的联系更紧密。作家莫莉·菲舍尔（Molly Fischer）在2020年的一篇文章中将其称为“千禧一代的审美”；它也受到了床垫销售商Casper和共享办公连锁企业WeWork和The Wing等初创公司的追捧。菲舍尔问道：“千禧一代的审美会终结吗？”

As years passed, however, I realized that AirSpace was less of a specific style than a condition that we existed in, something beyond a single aesthetic trend. Like all fashions, the visual style of that moment in the mid-2010s decayed. The white subway tiles began to look too cliché, the way the laminate countertops in my childhood home felt like the nineties, and they were replaced by brightly colored or more textured ceramic tile. The early rough-hewn style of high Brooklyn lumberjack, with its repurposed industrial furniture, gave way to careful Scandinavian-ish mid-century modernism, with spindly-legged chairs and wood joinery. In the late 2010s, the dominant aesthetic grew colder and more minimal, with cement countertops and harsh geometric boxes in place of chairs. Accoutrements like lights made from rusty plumbing fixtures were left behind in favor of houseplants (succulents especially) and highly textured fiber art, evoking West Coast bohemia more than hardscrabble New York City. The association with Brooklyn gradually faded out (after the 2020 pandemic, Brooklyn itself was seen as less desirable than downtown Manhattan), and the generic style became less associated with a place than with digital platforms, like Instagram and the insurgent TikTok. In a 2020 essay, the writer Molly Fischer labeled it “the millennial aesthetic”; it was also embraced by start-up companies like the mattress seller Casper and the coworking space chains WeWork and The Wing. Fischer asked, “Will the millennial aesthetic ever end?”

风格元素——爱迪生灯泡与霓虹灯——最终被证明不如根本的同质性重要，而后者却愈发根深蒂固。多年来，标识牌不断演变，一步步演变，但其同一性始终如一。即使在19世纪，塔德就曾预言，未来风格的差异将不再基于“空间的多样性”，而是“时间的多样性”。

The elements of style—Edison bulbs versus neon signs—turned out to be less important than the fundamental homogeneity, which became more and more entrenched. The signs changed, evolving one step at a time over the years, but the sameness stayed the same. Even in the nineteenth century, Tarde predicted that in the future stylistic difference would be based not on “diversity in space” but “diversity in time.”

每一种新的审美转变都随之蔓延开来。在我以前居住的布鲁克林布什维克街区，一家名为“燕子”（Swallow）的昏暗工业风咖啡店，其家具常常看起来像跳蚤市场甩卖的商品，曾经引领时尚潮流，但后来被“超级皇冠”（Supercrown）超越。后者是一家更时尚的咖啡馆，配有配套的凳子和宽敞的天窗。“超级皇冠”关门后，几条街外的“赛伊咖啡”（Sey Coffee）出现了。远离喧嚣。这是一个简朴的环境，多肉植物——有些甚至嵌入墙壁——的数量比舒适的椅子还要多。裸露的砖墙被漆成了白色。咖啡馆的后方曾是一个完整的陶瓷工作室，制作着咖啡馆的杯子，散发着侘寂的气息。这不是一个让人流连忘返的空间，但它的美学设计堪称完美。你几乎要把它发到Instagram上，在明亮的日光下，拍下你手中的卡布奇诺在吧台抛光混凝土上流淌的瞬间。Sey 体现了这种为数字平台而生的、无摩擦、易于分享的美学。

Each new aesthetic shift became pervasive in turn. Within my old Brooklyn neighborhood of Bushwick, a dim, industrial-chic coffee shop called Swallow whose furniture often resembled a flea-market sale was once the height of fashion, but it was outpaced by Supercrown, a glossier space with matching stools and an expansive skylight. After Supercrown shut down, Sey Coffee emerged a few blocks away. It’s a spartan environment in which succulents—some even embedded in the walls—outnumber comfortable chairs. Its exposed brick walls have been painted white. At one point, its rear held an entire ceramics studio, which produced the café’s cups, redolent of wabi-sabi. It is not a space in which to linger, but it is aesthetically perfect. You almost have to Instagram it, capturing your cappuccino in bright daylight on the polished concrete of the bar. Sey embodied that frictionless, easily shared aesthetic developed for digital platforms.

有时人们会对咖啡店里出现的地铁瓷砖感到不满。但我认为，真正令人反感的是“空域”的千篇一律，而不是其风格本身。在一个多元化的世界里，同质化令人毛骨悚然。在另一个地方找到预期的审美可能会让人失望——一种日益增长的无聊感——同时也会产生一种被侵犯的感觉，数字平台的影响力正在扩展到它此前从未触及的领域，这是 Filterworld 扩张的标志。

Sometimes people expressed annoyance at the very presence of subway tiles in coffee shops. But I think it was the sameness of AirSpace that was off-putting, not exactly the style itself. Homogeneity in a diverse world is uncanny. There could be a disappointment with finding the expected aesthetic in yet another place—a burgeoning boredom—as well as a sense of intrusion, that the influence of digital platforms was extending somewhere that it had not theretofore reached, a sign of Filterworld’s expansion.

2010 年代末，一位名叫萨丽塔·皮莱·冈萨雷斯的南非女性在开普敦一家城市化非营利组织工作时，注意到了这种美学。冈萨雷斯认为这是一种绅士化，甚至是后殖民国家殖民主义的回响。在开普敦中心区的克鲁夫街上，大量极简主义风格的咖啡馆应运而生。在我与冈萨雷斯交谈时，她一一列举了这些咖啡馆的特点，并指出这些咖啡馆“有着长长的木桌、锻铁装饰、悬挂的灯泡和吊挂的植物”。这种美学本身也蔓延到了不同的场所：啤酒馆、美食酒吧、艺术画廊、爱彼迎。“不仅仅是咖啡馆；即使是那些出租或翻修历史建筑的人，很多人也在遵循这种美学，”冈萨雷斯说。 2016 年左右，她住在明尼阿波利斯东北部时，注意到那里也发生了类似的变化，那里的仓库建筑变成了咖啡店、小型啤酒厂和联合办公室，这些都是高档社区的常见指标。

A South African woman named Sarita Pillay Gonzalez noticed the aesthetic in Cape Town in the late 2010s, when she was working there at an urbanism nonprofit. Gonzalez saw it as a form of gentrification, or even an echo of colonialism in a postcolonial country. Generically minimalist coffee shops were popping up on Kloof Street, in Cape Town Central. Gonzalez identified them by their “long wooden tables, wrought-iron finishings, those lightbulbs that hang, hanging plants,” she reeled off the list when I spoke with her. The aesthetic itself was propagating into different venues as well: beer halls, gastropubs, art galleries, Airbnbs. “It’s not just coffee shops; even people who are renting out or renovating historical homes, a lot of them are also hewing to that aesthetic, too,” Gonzalez said. She had noticed a similar transformation in northeast Minneapolis while she was living there around 2016, where warehouse buildings were turned into coffee shops, microbreweries, and coworking offices, the common indicators of a gentrifying neighborhood.

冈萨雷斯认为，这种风格标志着“一个全球都能接触到的空间。你可以从曼谷到纽约，从伦敦到南非，再到孟买，都能找到同样的感觉。你可以轻松地融入那个空间，因为它是一个如此熟悉的空间。”同质化与2010年代普遍的嬉皮士哲学形成了鲜明对比。嬉皮士哲学认为，通过消费某些产品和文化产品，你可以彰显自己与主流人群的独特之处——在这里，你指的是一家特定的咖啡店，而不是一个默默无闻的乐队或服装品牌。“讽刺的是，这些空间本应代表个性，但它们却极其单调，”冈萨雷斯说道。她的评论呼应了Filterworld和算法推荐的另一个自相矛盾的信息：你是独一无二的，就像其他人一样。

According to Gonzalez, the style marked “a globally accessible space. You’re able to hop from Bangkok to New York to London to South Africa to Mumbai and you can find that same feel. You can ease into that space because it’s such a familiar space.” The homogeneity contrasted with the general hipster philosophy of the 2010s, namely, that by consuming certain products and cultural artifacts you could proclaim your own uniqueness from the mainstream crowd—in this case a particular coffee shop rather than an obscure band or clothing brand. “The irony of it all is that these spaces are supposed to represent spaces of individuality, but they’re incredibly monotonous,” Gonzalez said. Her comment echoed another paradoxical message of Filterworld and algorithmic recommendations: You are unique, just like everybody else.

冈萨雷斯观察到，不仅空间，就连顾客也都千篇一律：“如果你走进咖啡馆，你会发现他们以白人为主。但从历史上看，这里曾是一个有色人种聚居的社区。这无疑与士绅化阶层息息相关。” 只有特定类型的人被鼓励在“空域”区域感到舒适自在，其他人则被主动过滤掉。士绅化也是一种扁平化——这一点尤其体现在那些占据着廉价社区的翻新建筑的美学设计上：砖块被漆成灰色，木质栏杆变成了拉力钢丝，无衬线的门牌号被贴在门边。要适应这种典型的行为，需要一定的金钱和一定的熟练程度，比如把笔记本电脑随意地放在普通咖啡馆的宽桌上，一坐就是几个小时，就像学习豪华酒店鸡尾酒吧里不成文的礼仪一样。冈萨雷斯说，AirSpace 咖啡馆“令人感到压抑，因为它们排外又昂贵”。当白人和富人被视为常态时，一种美学和意识形态的力场会将任何不符合这一模式的人拒之门外。

Not only the spaces but the customers were homogenous, Gonzalez observed: “If you go into the cafés, they’re predominantly white. But it’s historically a neighborhood for people of color. It’s definitely associated with the gentrifying class of people.” Only certain types of people were encouraged to feel comfortable in the zone of AirSpace, and others were actively filtered out. Gentrification is a form of flattening, too—a fact visible not least from the aesthetics of renovated buildings that take over affordable neighborhoods, the bricks painted gray, the wooden railings turned into tension wire, the sans serif address numbers mounted next to the door. It required money and a certain fluency for someone to be comfortable with the characteristic act of plunking down a laptop on one of the generic cafés’ broad tables and sitting there for hours, akin to learning the unspoken etiquette of a cocktail bar in a luxury hotel. The AirSpace cafés “are oppressive, in the sense that they are exclusive and expensive,” Gonzalez said. When whiteness and wealth are posed as the norm, a kind of force field of aesthetics and ideology keeps out anyone who does not fit the template.

平坦理论

THEORIES OF FLATNESS

21世纪初的美国，我从小就接受“世界是平的”这一观念。那几年，全球化的主流意识逐渐萌芽，殖民主义和资本主义导致地球比以往任何时候都更加互联互通，也让人感觉更加渺小。促成这一流行观念的主要原因是《纽约时报》专栏作家托马斯·弗里德曼2005年出版的《世界是平的》。这本书当时被认为是一种普遍共识：平坦意味着人、商品和……思想在物理空间中的流动比以往任何时候都更快、更容易。那是历史上一个动荡的时刻，但即使是9/11事件以及随后持续不断的战争，也让人们深刻地认识到，美国与地球其他地区并非如此遥远或分离。“世界是平的”因此变成了一个矛盾的教训：你可以消费大量的中国制造产品，但中国发生的事情也可能影响到你个人。

In the early 2000s United States, I grew up with the idea that the world was flat. Those years saw the dawning mainstream awareness of globalization, the way that colonialism and capitalism had led to a planet that was more interconnected and felt smaller than ever before. The major culprit for this popular idea was the New York Times columnist Thomas Friedman’s 2005 book The World Is Flat. It felt like common wisdom: flatness meant that people, goods, and ideas flowed across physical space faster and easier than ever. It was a turbulent moment in history, but even 9/11, and the enduring wars that followed, drove home a certain visceral lesson that America wasn’t so distant or separate from the rest of the planet. “The world is flat” thus became an ambivalent lesson: You can consume plentiful products manufactured in China, but what happens in China might also affect you personally.

弗里德曼在书中探讨了各种“扁平化因素”，这些因素正在将地球紧密地联系在一起。其中一些因素是数字技术：像网景（Netscape）这样价格实惠的互联网浏览器；促进跨国企业和工厂之间协作的工作流软件；以及像谷歌（Google）这样扩展信息获取渠道的搜索引擎。弗里德曼写道：“地球历史上从未有过如此多的人——独自一人——能够找到如此多关于如此多事物和如此多其他人的信息。” 这本书主要探讨的是国家和企业（它们正在迅速变得更加相似）的宏观层面。“竞争环境”正在变得公平，这意味着任何公司无论位于何处都可以与其他公司竞争。等级制度正在被扁平化，因此小公司或自由职业者可以与规模大得多的企业竞争。正如高速公路连接美国一样，互联网的光纤网络创造了“一个更加无缝的全球商业网络”，弗里德曼写道，“并有助于打破全球区域主义”。

In his book, Friedman wrote about various “flatteners,” forces that were knitting the planet closer together. Several of these were digital technology: affordable Internet browsers like Netscape; workflow software that enabled collaboration between international businesses and factories; and search engines like Google, which expanded access to information. “Never before in the history of the planet have so many people—on their own—had the ability to find so much information about so many things and about so many other people,” Friedman wrote. The book mostly addressed the macroscopic level of nations and corporations (which were rapidly becoming more similar). The “playing field” was being leveled, which meant that any company could compete against any other no matter where they were located. Hierarchies were being flattened, so a small company or a freelancer could compete with much larger enterprises. Just as highways interconnected the United States, the fiber networks of the Internet created “a more seamless global commercial network,” Friedman wrote, and “helped to break down global regionalism.”

在新的全球化秩序下，不仅产业和经济被扁平化，文化也呈现出同样的趋势。新兴的互联网带来了共享的压力，它在微观层面上连接了个人，就像国家和企业连接一样。“每个人都希望所有东西都尽可能数字化，这样他们就可以通过互联网管道将其发送给其他人，”弗里德曼写道。从那时起，这些管道变得越来越大，速度也越来越快。

Not only were industries and economies being flattened in the new globalized order, but culture trended that way as well. The nascent Internet exerted a pressure to share, and it connected individuals on a microscopic level in the same way that countries and corporations were being connected. “Everyone wanted everything digitized as much as possible so they could send it to someone else down the Internet pipes,” Friedman wrote. The pipes have since grown much larger and faster.

过滤世界是弗里德曼平坦世界的终点，因为相同的力量正日益影响着我们个人生活的方方面面，直至潜意识层面。除了关于制造订单或消费者需求的枯燥统计数据之外，图像和视频照片开始以前所未有的速度在全球流动，尤其是在社交媒体的兴起之后，社交媒体鼓励用户创作自己的内容。社交网络在弗里德曼出版这本书之后才开始崭露头角，但与网景公司类似，它们实现了在线数字媒体的民主化访问——使创作和消费都成为可能。Flickr 于 2004 年推出，用户可以通过一个基本的社交网络在个人账户上上传、发布和整理照片，其目标用户是业余摄影师。YouTube 成立于 2005 年，允许任何拥有足够强大互联网连接的人上传和分享视频片段。Instagram 紧随其后于 2010 年推出，并创造了一种更广泛的分享新近成为主流的 iPhone 相机快照的文化。这些也随之全球化。

Filterworld is the end point of Friedman’s flat world, because the same forces are increasingly acting on every aspect of our personal lives, down to the level of the subconscious. Beyond dry statistics about manufacturing orders or consumer demand, images and video began flowing across the world faster than ever, particularly with the rise of social media, which encouraged users to generate their own content. Social networks only came to the fore after Friedman’s book, but similar to Netscape, they democratized access to online digital media—making it possible to create it as well as consume it. Flickr, which gave users the ability to upload, publish, and organize photographs on individual accounts with a rudimentary social network, launched in 2004 and targeted amateur photographers. YouTube was founded in 2005. It allowed anyone with a powerful enough Internet connection to upload and share video clips. Instagram followed in 2010 and created a larger culture of sharing snapshots from newly mainstream iPhone cameras. These became globalized, too.

没有什么比《江南Style》更能体现这种转变了。这部由韩国说唱歌手Psy执导、赵秀贤执导的音乐录影带于2012年夏天在YouTube上首播。这段时长四分钟的MV从一开始就是一部国际化的杰作，它打破了美国MTV确立的格式、欧洲俱乐部电子乐的合成器音色以及韩国流行音乐产业在亚洲的流行趋势。但这部MV也极具特色，它以一首主要用韩语演唱、讽刺首尔最富裕社区的歌曲为基础。（除了那句史诗般的英文副歌：“Eh, sexy lady.”）2012年12月，《江南Style》成为YouTube上首个点击量突破10亿的视频。最终，它的点击量超过了40亿。

Nothing epitomized the shift more than “Gangnam Style,” the music video from the South Korean rapper Psy, directed by Cho Soo-hyun, that premiered on YouTube in the summer of 2012. The four-minute-long video was from the outset an internationalized artifact, emerging from the format established by American MTV, the synthesizer palette of European club techno, and the pan-Asian popularity of the Korean pop music industry. But the video was also deeply specific, building on a song that satirized the wealthiest neighborhood in Seoul, sung primarily in Korean. (Except for the epic English-language refrain: “Eh, sexy lady.”) In December of 2012, “Gangnam Style” became the first YouTube video to pass a billion views. It eventually exceeded four billion.

当这首歌和视频流行起来时，我记得自己当时感到很惊讶，因为我可以如此轻松地欣赏韩流音乐，就像观看任何美国音乐视频一样流畅。多亏了YouTube，这种体验才如此顺畅，部分原因也归功于该平台的算法推荐，才有这么多人观看。（记住，在Filterworld，关注会引发关注。）2012年，YouTube还将“观看时长”变量整合到其算法中，这意味着“你观看了哪些视频以及观看了多长时间”，据该公司的一篇博客文章称，这带来了更有针对性的推荐。突然之间，通过数字平台，全世界大部分地区都在观看同一件事。而这仅仅是数字全球化的早期阶段。十年后的今天，超过 30 个 YouTube 视频的观看次数已超过 30 亿——这象征着 YouTube 在多大程度上吸引了电视观众的注意力，以及其用户群有多广泛。

When the song and video became popular, I remember feeling bemused at how easily I could consume K-Pop, as smoothly as watching any American music video. It was thanks to YouTube that the experience was so frictionless, and due in part to the platform’s algorithmic recommendations that so many people saw it. (Remember that in Filterworld, attention begets attention.) Two thousand and twelve was also when YouTube integrated the variable of “watchtime” to its algorithm, meaning “which videos you watched and for how long,” according to a company blog post, leading to better-targeted recommendations. Suddenly, via digital platforms, much of the world was watching the same thing. And that was just the early days of digital globalization. Now, a decade later, there are over thirty YouTube videos with over three billion views—a symbol of just how much YouTube has absorbed audiences’ attention away from TV as well as how widespread its user base has become.

全球化一词通常指智能手机等商业产品的普及，民主等政治理念，以及对美国主导的伊拉克战争等国际冲突的干预。但这种互联互通也导致个人体验更加平凡和普遍地扁平化。在美国，我和印度、巴西或南非的互联网用户使用相同的设备，访问许多相同的社交网络，并连接到相同的流媒体服务。弗里德曼的国际竞争加剧模型导致只有少数赢家，它们从对国际化数字空间的垄断中获利颇丰。（个人和小型企业之间竞争，但主要是为了在规模更大的平台公司格局中购买商品，就像狼群争夺地盘一样。）

Globalization is often used as a term in relation to the availability of commercial products like smartphones; political ideas like democracy; and intervention in international conflicts like the American-led Iraq War. But such interconnection has also led to a more mundane and pervasive flattening of individual experiences. In the United States, I use the same devices, access many of the same social networks, and connect to the same streaming services as an Internet user in India, Brazil, or South Africa. Friedman’s model of increased international competition has resulted in only a few overall winners, which profit hugely from their monopolization of the internationalized digital space. (Individuals and small businesses compete, but mainly with each other for purchase in the landscape of much larger platform companies, like wolves competing for territory.)

事实上，早在《世界是平的》问世十多年前，文化理论家们就已开始描述全球化，尤其是互联网加速的全球化，是如何滋生千篇一律和千篇一律的。人们对这种千篇一律的焦虑日益加剧，对全球化带来的文化后果感到不满。1989年，西班牙社会学家曼努埃尔·卡斯特尔构想了“流动空间”，并在1999年的一篇文章中将其定义为“允许社会实践在不受地域限制的情况下同时进行的物质安排”。换句话说，卡斯特尔认为，像互联网这样的电子电信基础设施使得共享文化能够跨越距离而共同发展，而不是依赖于物理上的接近性。在电信网络覆盖的不同地区，所形成的文化是相同的。这与卡斯特尔称之为“地方空间”的物理地理学背道而驰。

In fact, for more than a decade before The World Is Flat, cultural theorists were already describing how globalization, particularly the accelerated version caused by the Internet, breeds sameness and monotony. There was also a rising anxiety about this sameness, a disaffection with the cultural consequences of globalization. In 1989, the Spanish sociologist Manuel Castells conceived of the “space of flows,” which he defined as “the material arrangements that allow for simultaneity of social practices without territorial contiguity,” as he wrote in a 1999 article. In other words, Castells argued, electronic telecommunications infrastructure like the Internet enabled shared culture to be collectively developed across distances rather than relying on physical proximity. The culture that formed was the same across the disparate places that the telecommunications networks covered. It was a departure from physical geography, which Castells labeled the “space of places.”

卡斯特写道，人类社会的许多不同方面开始更多地遵循流动空间的逻辑，而非地点空间的逻辑，包括“金融市场、高科技制造、商业服务、娱乐、媒体新闻、毒品贩运、科学技术、时装设计、艺术、体育或宗教”。地理的重要性因此进一步消退。2001年，卡斯特写道，流动空间“围绕共同的功能和意义，将遥远的地点连接起来……同时孤立和压制了地方空间所蕴含的经验逻辑。” 无论对于你的体验还是你的亲和力而言，你身处何方，都不如你消费的媒体渠道重要。

Many different aspects of human society began to follow the logic of the space of flows more than the space of places, including “financial markets, high-technology manufacturing, business services, entertainment, media news, drug traffic, science and technology, fashion design, art, sports, or religion,” Castells wrote. The significance of geography thus receded further. In 2001, Castells wrote that the space of flows “links up distant locales around shared functions and meanings…while isolating and subduing the logic of experience embodied in the space of places.” Where you were physically mattered less, both to your experiences and your affinities, than which channels of media you were consuming.

如果说地理因素的重要性正在下降，那么交通和活动区域则更加重要。1992年，法国哲学家马克·奥热（Marc Augé）撰写了一本名为《非场所》（Non-Places）的书，研究了高速公路、机场和酒店的感官体验：这些区域在世界各地都变得非常相似。它们给现代游牧民带来了一种独特而矛盾的舒适感，他们属于“无场所”地带。在非场所，“人们总是，又永远不会在家，”奥热写道。这本书的引言讲述了一位法国商人开车前往戴高乐机场，快速通过安检，在免税店购物，然后在登机前等候。机场这个毫无个性的空间获得了

If geography was becoming less significant, then zones of transportation and movement mattered more. In 1992, the French philosopher Marc Augé wrote a book titled Non-Places, which studied the sensory experiences of highways, airports, and hotels: zones that had become reliably similar the world over. They lent a distinct, paradoxical sense of comfort to the modern nomad, who belonged to the placeless zone. In non-places, “people are always, and never, at home,” Augé wrote. The book’s introduction narrates a French businessman driving to the Charles de Gaulle airport, zipping through security, shopping duty-free, and then waiting before the plane boards. The personality-less space of the airport attains

荒地、庭院和建筑工地、旅客休息的车站台和候车室，以及所有偶然相遇的地方所蕴含的不确定的魅力，在这些地方，人们会产生继续冒险的可能性的逃避感。

something of the uncertain charm of the waste lands, the yards and building sites, the station platforms and waiting rooms where travelers break step, of all the chance meeting places where fugitive feelings occur of the possibility of continuing adventure.

登机的流程，以及随后令人麻木的飞行体验，都包含着一种自我与周遭环境的剥离，直到一切都变得顺畅一致。这是一种显而易见的感觉——飞机起飞时那种与现实的微微分离，或是第一次打开酒店房门时那种彻底的匿名感。“非场所空间既不会创造单一的身份，也不会创造关系；只有孤独和相似，”奥热写道。他描述了“身份丧失带来的被动的快乐”。就连这位虚构的商人在飞机上阅读的杂志，也提到了“国际商业环境中”“需求和消费模式的同质化”。

The procession to the flight and then the numbing experience of flying itself involves a kind of stripping-away of the self and surroundings until everything becomes smooth and uniform. It’s a recognizable feeling—that slight separation from reality that happens when the plane takes off, or the clean burst of anonymity when opening the door of a hotel room for the first time. “The space of non-place creates neither singular identity nor relations; only solitude and similitude,” Augé writes. He describes “the passive joys of identity-loss.” Even the magazine the fictional businessman reads on the plane references “the homogenization of needs and consumption patterns” in the “international business environment.”

在 20 世纪 90 年代和 21 世纪，人们对全球化的进程和这种空灵化抱有一种乐观主义或乌托邦主义的心态。经验。如果地球联系更紧密，有了新的机场和国际连锁酒店，或许人们就能更好地理解彼此。扁平化的地理布局就像世界语这种通用语言：它或许代表着一种简化，但至少每个人都会熟悉它。有人认为，千篇一律更有效率，这不仅是为了舒适（至少对富裕的西方旅行者而言），也是为了在资本以投资或基础设施的形式从一个地方流向另一个地方时，更好地滋养资本。既定模式更可预测，也能更快地扩大规模。混乱的多样性无利可图，就像个人品味会导致数字消费效率低下一样。

In the 1990s and 2000s, there was a kind of optimism or utopianism to the march of globalization and this etherealization of experience. If the planet was brought closer together, with new airports and international hotel chains, perhaps people would understand each other better. The flattening of geography was something like the universal language of Esperanto: it might represent a simplification, but at least it would be familiar to everyone. There was an argument to be made that sameness was more efficient, not only for comfort’s sake (at least for the wealthy Western traveler) but for the nurturing of capital as it flowed from one place to another in the form of investments or infrastructure. Set patterns were more predictable and could scale up faster. Chaotic diversity was unprofitable, just as personal taste makes for inefficient digital consumption.

荷兰建筑师雷姆·库哈斯深谙这种国际融合之道。九十年代，他的公司大都会建筑事务所（OMA）在欧洲各地开展业务，以其大胆的概念建筑而闻名，这些建筑并非最终得以建成。库哈斯的设计超凡脱俗。OMA 于 1996 年设计的“超级建筑”看起来像几座摩天大楼以锐角相互碰撞，可以建在任何地方，成为一座“可容纳 12 万人的自给自足的城市”，公司对此的描述是：“该建筑的结构隐喻着城市：塔楼构成街道，水平元素是公园，体量是区域，对角线是林荫大道。” 建筑的无个性，以及它与蓬勃发展的非场所领域的轻松融合，是设计中刻意为之的一部分——一种包容通用性的美学。库哈斯在1995年发表的一篇题​​为《通用城市》的文章中阐述了他的哲学。这篇文章是那种读过一次就念念不忘的短篇作品。它对美学和建筑理论的犀利阐述在21世纪被证明具有预见性。文章开头写道：“当代城市是否就像当代机场一样——‘千篇一律’？” “如果这种看似偶然——且通常令人遗憾——的同质化是一个有意为之的过程，一个有意识地从差异走向相似的过程呢？如果我们正在见证一场全球性的解放运动：‘打倒个性！’呢？”

The Dutch architect Rem Koolhaas was fluent in this international convergence. His firm, Office for Metropolitan Architecture, worked across Europe in the nineties and became known for daring conceptual buildings that weren’t always actually built. Koolhaas’s designs were otherworldly. OMA’s 1996 Hyperbuilding, which looks like several skyscrapers crashing together at sharp angles, could be built anywhere as a “self-contained city for 120,000,” as the firm’s description ran: “The building is structured as a metaphor of the city: towers constitute streets, horizontal elements are parks, volumes are districts, and diagonals are boulevards.” The characterlessness of the architecture, its easy merger into the burgeoning realm of non-places, was a purposeful part of the design—an aesthetic that embraces the generic. Koolhaas described his philosophy in a 1995 essay titled “The Generic City,” which is one of those short pieces of writing that you read once and it never leaves your mind. Its sharp declarations of aesthetic and architectural theory have proven prophetic in the twenty-first century. “Is the contemporary city like the contemporary airport—‘all the same’?” the essay begins. “What if this seemingly accidental—and usually regretted—homogenization was an intentional process, a conscious movement away from difference toward similarity? What if we are witnessing a global liberation movement: ‘down with character!’ ”

库哈斯的“通用城市”是所有城市居民居住的地方，有高档的AirSpace大楼和咖啡馆；有为科技工作者服务的共享办公空间；以及同样的餐厅和酒吧。这是我们从……下来后到达的熟悉的地方。飞机，穿过机场，然后乘车前往阁楼式酒店，在那里我们用手机办理入住手续。库哈斯认为，国际上身份认同的扁平化可以是积极的，或者至少有其自身的优点：“身份认同越强，它就越禁锢，就越抵制扩张、诠释、更新和矛盾。”以巴黎为例：“巴黎只会变得更具巴黎特色——它已经在走向超级巴黎，一个精致的漫画。”一个地方的独特性只会吸引更多的游客，而越来越多的游客会逐渐把它碾成尘土，游客来到这里将其特色消费成产品，却让它变得越来越堕落。差异只会成为障碍；它在这个越来越无摩擦的世界中制造摩擦，无论是在城市中还是在音乐中。正如波兰社会学家齐格蒙特·鲍曼在2000年左右提出的“液态现代性”概念所言，流动的能力是新时代权力的关键：“统治者是最难以捉摸的，那些可以自由流动而不被察觉的人。” 同样的情况也发生在过滤世界中，对于那些愿意接受算法在特定时刻优先考虑的内容和美学的人来说，他们不受任何身份的束缚。

Koolhaas’s generic city is the place that all urban residents inhabit, with gentrified AirSpace buildings and coffee shops; coworking spaces serving tech workers; and the same set of restaurants and bars. It is the familiar place we arrive at, stepping off a plane, rolling through an airport, and taking a car to a loft-style hotel, where we check in using our phone. Koolhaas poses the international flattening of identity as something that can be positive, or at least include its own advantages: “The stronger identity, the more it imprisons, the more it resists expansion, interpretation, renewal, contradiction.” Take Paris, for example: “Paris can only become more Parisian—it is already on its way to becoming hyper-Paris, a polished caricature.” A place’s uniqueness only attracts more tourists, which gradually grind it into dust with the increasing flow of travelers, who arrive to consume its character as a product and leave it ever more degraded. Difference just gets in the way; it creates friction in a world that is increasingly frictionless, whether in its cities or in its music. The ability to flow is the key to power in the new era, as the Polish sociologist Zygmunt Bauman observed with his circa-2000 concept of “liquid modernity”: “It is the most elusive, those free to move without notice, who rule.” The same is true in Filterworld for those who are willing to take up whichever content and aesthetics algorithms are prioritizing in a given moment, not anchored by any one identity.

我们必须带着某种讽刺和对荒诞的包容来解读库哈斯的写作。这位建筑师有着艺术家的倾向，喜欢发表一些宏大的宣言，但这些宣言往往缺乏事实或数据的支持。他的文章最好被解读为一种挑衅，一种关于未来建筑如何运作或人们如何生活在其中的简短而狂喜的愿景。库哈斯认为，即使从1995年的视角来看，互联网也导致了地方认同感的衰落：“通用城市是城市生活的大部分跨越到网络空间后剩下的东西，”他写道。“它是一个感觉虚弱而膨胀的地方，情感稀少，如同床头灯照亮的广阔空间，谨慎而神秘。” （这句话让人想起现在普遍存在的一种体验：在昏暗的卧室里盯着手机，滚动浏览信息。）在这个描述中，有曼努埃尔·卡斯特尔斯 (Manuel Castells) 解释流动空间如何从地点空间中抽走意义的痕迹，因为更多的生活和文化发生在物理位置之间或跨物理位置，而不是在物理位置内部。

One must take Koolhaas’s writing with a degree of irony and a tolerance for the absurd. The architect has an artist’s tendency to make grand pronouncements that he doesn’t entirely back up with facts or data. His essays are best read as provocations, brief, ecstatic visions of how architecture might work or how people might live in it in the future. Koolhaas saw the decline of local identity as a consequence of the Internet, even from the perspective of 1995: “The Generic City is what is left after large sections of urban life crossed over to cyberspace,” he wrote. “It is a place of weak and distended sensations, few and far between emotions, discreet and mysterious like a large space lit by a bed lamp.” (The line evokes the now-universal experience of staring at your phone in a dim bedroom, scrolling through feeds.) In this description, there are shades of how Manuel Castells explained the space of flows draining meaning away from the space of places, as more of life and culture took place between or across physical locations than within them.

普通城市或流动空间，或扁平的世界，逐渐创造出自己的语境，拥有自己的规范和期望。它“引发了一种对常态的幻觉，”库哈斯写道。之所以说是“幻觉”，是因为它并非纯粹的有机体，而是一种由科技引发的幻象，如同一场狂热的梦境；之所以说是“常态”，是因为它是一种同质化的模板，一种重复的模式，其普遍性确立了自身的常态。千篇一律的城市已经肆无忌惮地蔓延，不受控制。

The generic city or the space of flows, or the flattened world, gradually creates its own context with its own norms and expectations. It “induces a hallucination of the normal,” Koolhaas wrote. A “hallucination” because it isn’t purely organic, it is a vision induced by technology, like a fever dream, and “normal” because it is a homogenized template, a repeating pattern whose ubiquity establishes its own normalcy. The generic city has spread implacably, unchecked.

盖亚特里·查克拉沃蒂·斯皮瓦克是一位文学理论家，1942年出生于印度加尔各答，被认为是后殖民理论的先驱之一。她来自西方之外，但在康奈尔大学和剑桥大学等西方名校接受教育，这让她以独特的视角审视和批判二十世纪余波。“1989年后，资本主义的胜利引领了全球化，”她在2012年出版的文集《全球化时代的审美教育》中写道。她写道，通过仅仅从经济生产力的角度来衡量一切，资本主义已经将生活的许多方面变成了“近乎完全的抽象”。其后果之一就是“全球化令人麻木的千篇一律”。用托马斯·弗里德曼的话来说，世界是平的，但扁平化也暴露了其自身的沉闷。

Gayatri Chakravorty Spivak is a literary theorist who was born in Calcutta, India, in 1942 and is considered one of the pioneers of postcolonial theory. Coming from outside of the West but educated in Western institutions like Cornell University and Cambridge, hers is a unique prism through which to survey and critique the aftermath of the twentieth century. “After 1989, capitalism triumphant has led through to globalization,” she wrote in her 2012 collection An Aesthetic Education in the Era of Globalization. By evaluating everything solely in terms of financial productivity, capitalism has turned so many aspects of life into “nearly complete abstraction,” she wrote. One consequence is “the mind-numbing uniformization of globalization.” The world is flat, per Thomas Friedman, but flatness has revealed itself to be stultifying.

我在本书的引言中引用了斯皮瓦克的一句名言：“全球化只发生在资本和数据领域。其他一切都是为了控制损失。” 我们谈论政治、文化和旅行的全球化，但从更根本的层面来看，斯皮瓦克的观点是正确的：真正在全球流动的是各种形式的金钱和信息：投资、企业、基础设施、服务器集群，以及所有数字平台的综合数据，它们像风或洋流一样在国家之间无形地流动。我们用户自愿通过同一个系统传输自己的信息，也将自己变成了流动的商品。

I referenced one of Spivak’s aphorisms in this book’s introduction: “Globalization takes place only in capital and data. Everything else is damage control.” We talk about politics, culture, and travel becoming globalized, but on a more fundamental level, Spivak is correct that what really flows across the planet are various forms of money and information: investments, corporations, infrastructure, server farms, and the combined data of all the digital platforms, sluicing invisibly like wind or ocean currents between nations. We users voluntarily pump our own information through this same system, turning ourselves into flowing commodities, too.

扁平化的历史意义重大，因为它表明扁平化本身就存在历史。过滤世界的同质化并非我们当下才有的现象；它是算法推送出现之前很久就发生的变化的必然结果，而且未来很可能还会加剧。毕竟，每次宣布一个宏大的扁平化趋势，世界总会以某种方式变得更加扁平。

This history of flatness is important because it shows that flatness has a history. The homogenization of Filterworld is not just a phenomenon of our own moment; it is a consequence of changes that happened long before algorithmic feeds and is just as likely to intensify in the future. After all, each time a grand flattening is announced the world somehow finds a way to get even flatter.

奥热、库哈斯和斯皮瓦克等思想家都引用软件作为比喻，来描述地理位置和国家是如何形成的。在全球互联互通的时代，人们试图使自己变得相似，这也是导致这种同质化的一个因素。但在社交媒体时代，同​​样的效果也发生在个人层面，包括文化消费者和文化创造者，他们都登录着同一套应用程序。取代实体酒店和机场的是 Twitter、Facebook、Instagram 和 TikTok，它们是消除差异的聚集空间。超越库哈斯的“通用城市”，现在还有通用的全球消费者，他们的偏好和欲望更多地受到他们使用的平台而不是居住地的影响。在某些情况下，我们的生活更多地是在流动空间而不是地点空间中进行。我们才刚刚开始理解，我们和我们的世界一样，现在也是平的。

Thinkers like Augé, Koolhaas, and Spivak all cited software as a metaphor to describe how geographical places and nations came to resemble each other in an era of overwhelming global interconnection, and as a factor in how such homogenization happened. But in the era of social media, the same effect has occurred at the level of the person, including both cultural consumers and cultural creators, who all log in to the same set of apps. In the place of physical hotels and airports, we have Twitter, Facebook, Instagram, and TikTok as spaces of congregation that erase differences. Moving past Koolhaas’s “Generic City,” there is now also the generic global consumer, whose preferences and desires are molded more by the platforms they use than where they live. In some cases, we conduct our lives more in the space of flows than the space of places. We are only beginning to understand how, along with our world, we are now flat, too.

作为一名记者，我经常在网上闲逛，充当着文化内容的筛选者，我是这个体系的参与者和推动者。我并非特别喜欢它，也并非希望这种同质化进一步蔓延。但大多数“过滤世界”的居民要么不愿，要么不知情。仅仅为了谋生或娱乐，我们就加速了这种扁平化。

As someone who hangs out online and acts as a sieve for culture-as-content in my career as a journalist, I am a participant and an accelerant of this system. It’s not that I particularly enjoy it or that I want the homogenization to spread further. But most of Filterworld’s inhabitants are either unwilling or unwitting. Simply by trying to make a living or entertain ourselves we accelerate the flattening.

普通咖啡馆老板

GENERIC CAFÉ OWNERS

2019年，我为了第一本书在京都进行研究旅行，在参观寺庙的假山花园之余，我也顺便体验了京城的咖啡馆风情。在20世纪初的日本，一种名为“喫茶店”的文化兴起：安静的茶室供应咖啡，不提供酒精饮料，以满足寻求宁静环境的作家和知识分子群体的需求。咖啡直到19世纪末才正式传入这个与世隔绝的国家，当时是由荷兰商人引入的。日本的咖啡馆以巴黎的咖啡馆为蓝本——尽管当时除了富人和开始阅读和翻译法国作家作品的学术知识分子外，很少有日本人会前往巴黎。

While I was traveling in Kyoto on a research trip for my first book in 2019, between visits to temple rock gardens I also traversed the city’s coffee-shop scene. In early-twentieth-century Japan, a culture of kissaten emerged: quiet tearooms that served coffee and excluded alcohol, catering to groups of writers and intellectuals seeking calm environments. Coffee had been officially imported to the isolated country only in the late nineteenth century, after it was first introduced by Dutch traders. The Japanese cafés were modeled on Parisian ones—though Paris was a place few Japanese people traveled to at the time, save the wealthy and the academic intelligentsia, who were beginning to read and translate French authors.

住在东京的一位朋友推荐我去六洋舍，这是一家1950 年在京都开业的茶餐厅。这家咖啡馆很难找到，它位于一条繁忙街道的地下楼梯下，一块不起眼的木牌子镶嵌在斑驳的绿松石色和棕色相间的方形瓷砖上，后来我才知道这些瓷砖是专门为这家店定制的。走进店里，仿佛进入了一个子宫：店内装饰着深色木质镶板，走廊大小的空间里摆放着皮革软垫长椅和双人卡座。店内一次最多只能容纳十个人。我从吧台后面经营这家店的老夫妇那里点了一杯手冲咖啡和一个自制甜甜圈，由于店内很安静，我说话的声音几乎低得像耳语。和其他客人一样，我坐在自己的卡座里，翻阅着一本书，在笔记本上记着什么。店里没有人看手机，我觉得部分原因是，这个地方似乎仍然保留着它最初建造时的模样。它鼓励人们关注细节和氛围，而不仅仅是用iPhone拍照所能捕捉到的。便携技术和互联网的非物质化时代与此格格不入。

I got a recommendation from a friend living in Tokyo to visit Rokuyosha, a kissaten-style café that opened in Kyoto in 1950. It was difficult to find, down basement steps on a busy street with an unobtrusive wooden sign mounted onto mottled turquoise-and-brown square ceramic tiles, which I later discovered were custom made for the shop. Descending into the shop itself was like entering a kind of womb: the interior was surfaced in ornamented dark wood paneling, with leather-padded benches and two-person booths lining the hallway-sized space. No more than ten people at a time could fit comfortably inside. I ordered a pour-over coffee and a homemade donut from the elderly couple who ran the place from behind the bar, using a voice not much above a whisper, since the café was silent. Like the other guests, I sat in my own booth, paged through a book, and wrote in a notebook. No one inside was looking at their phones, in part, I felt, because the space still seemed to exist in the time that it was originally built. It rewarded attention to the physical details and the atmosphere, beyond what could be picked up in an iPhone photo. The dematerialization of portable technology and the Internet just didn’t suit it.

六洋舍与我在京都其他地方找到的一家咖啡馆形成了鲜明对比。周末咖啡馆是我在谷歌地图上搜索咖啡馆时发现的；它在我的iPhone上用一个大点标记出来，因为我已经向Verizon支付了海外互联网费用，所以我仍然可以一直使用iPhone。（如果没有它，我会迷路的。）它位于市中心一个不起眼的停车场边缘——一个熙熙攘攘的空间——一栋两层小楼里。（它最初的地址是京都最早的意式浓缩咖啡店之一，于2005年开业。）建筑风格是传统的日式风格，推拉门，墙上抹着粗糙的灰泥，悬挂着宣纸球形灯。 （这种日式美学也被旧金山咖啡连锁店Blue Bottle所采用，它在每一座城市都布置了灯笼和插花风格花卉。2015年，Blue Bottle在东京开设了第一家略带复古风格的咖啡馆，一经推出便大受欢迎。）但Weekenders也带有AirSpace斯堪的纳维亚风格的影响，其开放式搁架和干净的金色木质台面也体现了这一点。正如马克·奥热（Marc Augé）所言，这样的咖啡馆是另一个让我们“永远，又永远不在家”的空间——一个国际化非场所中可识别的一部分。

Rokuyosha contrasted with a café I found elsewhere in Kyoto. Weekenders was a place I discovered on Google Maps by searching for cafés; it was marked by a large dot on my iPhone, which I could still use all the time because I had paid Verizon for Internet access abroad. (I would have been lost without it.) It was located at the edge of a nondescript parking lot in the city’s downtown—a space of constant movement—in a small two-story building. (Its original location opened as one of Kyoto’s first espresso shops in 2005.) The architecture was in a traditional Japanese mode, with sliding-panel doors, rough plaster on the walls, and hanging rice-paper globe lamps. (This Japanese aesthetic had also been adopted by the San Francisco coffee-shop chain Blue Bottle, which installed lanterns and ikebana-esque flower arrangements in every city it landed in. When Blue Bottle opened its first, slightly recursive Tokyo café in 2015, it was extremely popular.) But Weekenders also had hints of the AirSpace Scandinavian influence, with its open shelving and clean blond-wood countertops. Such a café is another space in which we are “always, and never, at home,” as Marc Augé wrote—a recognizable part of the international non-place.

我点了一杯卡布奇诺，当它摆在我面前时，我拿出手机拍下了瓷杯咖啡平衡地搁在柜台边上，柜台是一块长长的、粗糙的石头，像中世纪意大利教堂里的战利品一样。当然，我打算稍后将这张照片分享到网上。“人们来这里只是为了发 Instagram，”咖啡师用英语告诉我，语气中带着一丝干涩。我还记得自己拍这张照片时非常尴尬。我只不过是在故技重施。但这家咖啡馆低调的极简主义内饰和柔和的日光似乎正是它所需要的。现在查看 Weekenders 的 Instagram 帐户，多年来它标记的照片都是同一张自鸣得意的照片，也就是我拍的那张照片：一杯咖啡放在柜台上。

I had ordered a cappuccino, and when it was placed in front of me, I pulled out my phone to take a photo of the porcelain cup balanced on the edge of the counter, which was a long, roughened piece of stone, incorporated like spolia in a medieval Italian church. Of course, I intended to share the photo online later. “People come here just for the Instagram,” the barista told me in English, a slightly dry tone creeping in to his voice. I recall my acute sense of embarrassment taking the picture. I was performing a cliché. But with its unobtrusively minimalist interior and soft daylight, the café seemed to ask for it. Looking up Weekenders’s Instagram account now, its tagged photos over the course of years are a litany of the same self-satisfied image, the image that I took: a single coffee resting on the countertop.

简化为单一、原型、重复的图像并非偶然。这是一个漫长过程的终点。2010 年代初，出现了一种名为“Instagram 墙”的新现象。某种程度上，它是 21 世纪街头艺术运动的产物，涂鸦的高档化导致干净的、官方认可的壁画占据了城市墙壁，尤其是在破旧仓库林立的街区。街头艺术本身就成了一种景点，就像一个户外艺术画廊。我住在布什维克的时候，经常看到成群的法国游客在导游的带领下沿着工业区人烟稀少的街道走，仿佛那是卢浮宫，他们惊叹于壁画，而这些壁画最终被手绘的付费广告所取代。

The reduction to a single, archetypal, repetitive image wasn’t an accident. It was the end point of a longer process. In the early 2010s, a new phenomenon emerged called an “Instagram wall.” In part, it was an outgrowth of the street-art movement of the 2000s, a gentrification of graffiti that saw clean, officially sanctioned murals take over city walls, particularly in neighborhoods where decrepit warehouses were plentiful. Street art became an attraction in and of itself, like an outdoor art gallery. When I lived in Bushwick, I saw constant groups of French tourists being guided along the sparsely populated streets of the industrial neighborhood as if it were the Louvre, marveling at murals that eventually were replaced by hand-painted paid advertising.

街头艺术最初是一种游击活动，而 Instagram 墙则是专门为人们设计，让人们驻足拍照，然后发布到 Instagram 上的场所。它们也被称为“Instagram 陷阱”。有些只是色彩鲜艳的图形图案，为照片提供了完美的背景——墨西哥建筑师路易斯·巴拉甘 1948 年的故居，其柔和的粉红色墙壁就成了事实上的 Instagram 墙，吸引着游客。还有一些则创造了一个场景，让照片拍摄对象成为其中的一部分，类似于那些带有镂空图案的彩绘卡通木制道具，人们可以将脸探进去，假装成农民或足球运动员。Instagram 墙的典型代表，也是其最受欢迎的比喻之一，是一对天使般的翅膀在空旷空间的左右两侧展开，人们站在那里，经常伸展双臂，仿佛要飞翔。只需让朋友后退一步，拍照，然后发布即可！

While street art was originally a guerilla activity, Instagram walls were spots designed for people to stop and take photos in front of, to post on Instagram. They were also referred to as “Instagram traps.” Some were just bright-colored graphic patterns that provided a perfect backdrop for a photo—the lambent pink walls of the Mexican architect Luis Barragan’s 1948 home became a de facto Instagram wall, attracting tourists. Others created a scene that the photo subject became a part of, akin to those painted-cartoon wooden props with cutouts for people to poke their faces through and pretend to be a farmer or a football player. The epitome of the Instagram wall, one of its most popular tropes, was a pair of angelic wings unfurling to the left and right of an empty space where a person would stand, often stretching their arms upward as if taking flight. Just have a friend step back and take the photo, then post!

这种现象的巅峰或许是一家名为“迦太基必毁” (Carthage Must Be Destroyed) 的早午餐餐厅。这家餐厅于2017年在我当时居住的布什维克街区开业，地处一个遍布令人望而生畏的仓库的街区。餐厅内部空荡荡的——裸露的砖块和管道，公用木桌——但它却有一个独特的、大胆的设计噱头：所有东西都漆成了淡粉色。门是粉色的，柜台铺着粉色瓷砖，意式浓缩咖啡机的外壳也是粉色的，餐具则是釉面粉色的陶瓷餐具。菜单没什么特别之处，只是提供常见的吐司、牛油果等等（餐厅的联合创始人阿曼达·贝查拉 (Amanda Bechara) 是澳大利亚人），所以主要的吸引力在于它的美感。媒体照片一经发布，每个人都想去“那家粉色餐厅”。

The peak of this phenomenon might have been a brunch-focused restaurant called Carthage Must Be Destroyed. It opened in my then neighborhood of Bushwick in 2017, on a block full of forbidding warehouses. The interior was bare—exposed brick and plumbing, communal wood tables—but it had a single, aggressive design gimmick: Everything was painted pale pink. The door was pink, the counter was covered in pink tile, the espresso machine had pink housing, and the dishes were glazed-pink ceramic. The menu wasn’t particularly distinctive, offering the usual array of toasts, avocado and otherwise (its co-owner Amanda Bechara is Australian), so the main attraction was the aesthetic. The moment press photos were released, everyone wanted to go to “that pink restaurant.”

这个空间被优化为数字图像消费。当时，“千禧粉”——一种略带暗色调的腮红——因互联网而变得无处不在。它有时甚至被称为“Tumblr粉”，与它扎根的早期多媒体社交网络有关。耐克运动鞋、Glossier化妆品和Away行李箱上都能看到它的身影。苹果2015年发布的“玫瑰金”设备甚至也引领了这股潮流。Carthage餐厅堪称千禧粉体验的典范，一面沉浸式的Instagram墙。顾客们花了太多时间拍照，以至于餐厅制定了一项官方规定，禁止拍摄整个空间的快照——只能拍摄自己的食物。这项规定并没有真正发挥作用；它实际上要求顾客拍摄几张违规照片并发布。Instagram至今仍充斥着理论上违反规则的证据。

The space was optimized for consumption as a digital image. At the time, “millennial pink,” a slightly darkened blush color, had been made ubiquitous by the Internet. It was even sometimes known as “Tumblr pink,” associated with the early multimedia social network where it took root. It could be found on Nike sneakers, Glossier makeup products, and Away suitcases. Apple’s “rose gold” devices released in 2015 were even part of the trend. Carthage might as well have been the Millennial Pink Experience, an immersive Instagram wall. Visitors spent so much time taking photos that the restaurant had an official policy to disallow snapshots of the space as a whole—photos of your own food only. The policy didn’t really work; it all but demanded customers snap a few illicit pictures and post them. Instagram remains full today of evidence of theoretical rule breaking.

到了本世纪末，这些装置变得令人精疲力竭。所谓的 Instagram 博物馆应运而生，将拍照作为体验的刻意目标，仿佛去卢浮宫的唯一目的就是在蒙娜丽莎前自拍。甚至没有餐厅菜单来分散人们对数码图像制作的注意力。2017 年在旧金山开业的冰淇淋博物馆提供了以甜点为主题的沉浸式装置，同样在 2017 年开业的色彩工厂则提供了超现实的单色房间，用于展示戏剧性的肖像。它们都未能成为引人注目的视觉艺术，因为它们需要拍摄对象在场并拍照才能理解——除了这些数字平台并不完整；内容的制作才是最重要的。从这一点上来说，它们服务于一个非常明确的目的。Instagram 墙或体验吸引游客到某个地方，并通过让他们用手机进行活动来保持他们的参与度，比如餐厅为孩子们提供涂色书。这是对我们日益增长的成瘾性事实的让步——你不能随便去某个地方；你必须记录你的体验。当游客将这些照片发布到网上，最好是标记商家或地点时，这些照片就成了一种去中心化的在线广告牌，一种免费广告和数字口碑的形式。Instagram 墙就这样永存了下来。

These installations became utterly exhausting to encounter by the end of the decade. So-called Instagram Museums arose that made the taking of the photo the intentional goal of the experience, as if the only point of going to the Louvre was to take a selfie in front of the Mona Lisa. There wasn’t even a restaurant menu to distract from the act of digital image making. The Museum of Ice Cream that opened in San Francisco in 2017 offered dessert-themed immersive installations, and the Color Factory, also from 2017, surreal monochromatic rooms for dramatic portraits. Each failed as compelling visual art because they required the presence of the subject and the taking of a photograph to make sense—outside of digital platforms they were incomplete; the production of the content was all that mattered. In that, they served a very clear purpose. Instagram walls or experiences attracted visitors to a locale and kept them engaged by giving them an activity to perform with their phones, like a restaurant providing coloring books for kids. It was a concession to the fact of our growing addictions—you can’t just go somewhere; you must document your experience of it. And as visitors posted those photos online and ideally tagged the business or location, the photos became a kind of decentralized online billboard, a form of free advertising and digital word of mouth. The Instagram walls perpetuated themselves.

这些装置是搜索引擎优化的实体形式。Instagram 的墙面并非在网站上添加关键词，而是确保尽可能多的地点照片出现在数字平台上，从而扩大影响力。发布的帖子越多，出现的频率越高，推广算法就越能识别出该地点，并将其展示给更多潜在客户。这些墙面体现了一个迫在眉睫的事实：即使是实体场所，也必须像在现实生活中一样，在互联网上存在。

The installations were a physical form of search-engine optimization. Rather than including keywords on a website, the Instagram walls ensured that as many photos of a place as possible would exist on digital platforms, building a wider footprint. The more posts there were, and the more often they happened, the more promotion algorithms would also pick up on the place and display it to more potential customers. The walls spoke to the looming fact that even physical places have to exist as much on the Internet as they do in real life.

尽管植物墙已成陈词滥调，但它们的运作方式已渗透到空间和场所的方方面面，并开始优化我们所谓的“Instagram 化”。一家餐厅可能会有一面植物墙，墙上嵌着餐厅名称的霓虹灯招牌，从每张餐桌都能清晰可见，因此成为记录和分享的理想场所。一道菜品的视觉效果可能非常精致，以至于它更像是一幅图像，而非食物本身。纽约一家名为 Black Tap 的酒吧正是因此在 2016 年左右因其精心制作的奶昔而闻名。奶昔上点缀着糖果和其他装饰物（甚至还有一整块蛋糕），几乎无法食用，但在 Instagram 上发布时却显得格外引人注目。事实上，这些奶昔并非由厨师设计，而是由餐厅的社交媒体经理设计的。它们最初只在社交媒体网红的特别活动中供应，但后来越来越受到常客的欢迎，他们同样可以将它们转化为内容。它们更像是用来拍照的。比完成的还要多；过量的成分造成了不可持续的物理浪费。

Though the walls have become cliché, the way they work has been dispersed into every aspect of spaces and places, which began to optimize for what we called “Instagrammability.” A restaurant might include a living plant wall embedded with a neon sign of its name, easily visible from every table and thus an ideal target for documentation and sharing. A particular dish might be so elaborately visual that it functions more as an image than as food. That’s how a New York City bar called Black Tap became famous around 2016 for its elaborate milkshakes, which came so encrusted with candy and other accessories (even an entire piece of cake) that they were barely edible but looked appropriately dramatic in a photo for Instagram. In fact, the shakes were designed not by a chef but by the restaurant’s social media manager. They were first served only at special events for social media influencers, but became even more popular with regular customers, who could turn them into content just the same. They were more meant to be photographed than finished; the excess of ingredients created unsustainable physical waste.

站在京都的Weekenders咖啡馆，我感觉整家咖啡店都变成了一面Instagram墙，一个可以随时记录你作为消费者品味的标志性道具。上传照片也是一种展现你参与了这个令人向往的国际化“非场所”的方式，这个21世纪活力四射的创意旅行者的永恒家园。我很欣赏这种效果——毕竟，我追寻了那种千篇一律的体验，并且最终成功了。但同时也缺少了一些东西：旅行中，我并没有用陌生的事物给自己带来惊喜，只是通过在新地方发现它来重申自己味觉的优越性。或许这就是它感觉空洞的原因。

Standing at Weekenders in Kyoto, I felt that the entire coffee shop had become a kind of Instagram wall, a recognizable prop with which to document your tastes as a consumer. Posting a photo was a way of showing that you took part in the aspirational international non-place, the always-and-never home of the hypermobile twenty-first-century creative traveler. I appreciated the effect—after all, I had sought out that experience of sameness and achieved it. But there was also something missing: I wasn’t surprising myself with the unfamiliar during traveling, just reaffirming the superiority of my own sense of taste by finding it in a new place. Maybe that’s why it felt hollow.

科技对文化的影响往往潜移默化。它或许无处不在，又或许迅速彻底地改变着事物，以至于你忽略了其中的因果关系。它只是成为了新的现实。为了验证我对普通咖啡馆的直觉，我开始采访世界各地的一系列咖啡馆经营者，试图弄清楚他们为何要这样设计咖啡馆。他们中的许多人都认同AirSpace的工业极简主义美学。挪威咖啡先驱Tim Wendelboe于2007年开设了自己的第一家咖啡馆，他告诉我，这家咖啡馆标志性的北欧极简主义风格部分源于预算：他回收了自己建筑（前身是一家沙龙）中已有的材料，并用回收的木材建造了一个吧台。新加坡Chye Seng Huat Hardware的Xanthe Ang也提到了这家店的“五金传承”——它位于一栋装饰艺术时代的店屋内，拥有车库般的高挑天花板和金属灯具。马略卡岛 Mistral Coffee 的 Greg Schuler 指出，他的设计采用了“原始元素”：裸露的瓷砖地板、裸露的管道系统和胶合板架子。

The influence of technology on culture is often subtle. It can be so pervasive or change things so quickly and completely that you miss the connection between cause and effect. It simply becomes the new reality. To make sure my hunch about generic cafés was correct, I set out to talk to a series of coffee shop operators around the world to figure out why they designed their cafés the way they did. Many of them described the standard AirSpace aesthetic of industrial minimalism. Tim Wendelboe, the Norwegian coffee pioneer who opened his first café in 2007, told me that its signature Nordic minimalism was partly a consequence of budget: he recycled the material that was already in his building, a former salon, and used the reclaimed wood to build a bar counter. Xanthe Ang, of Singapore’s Chye Seng Huat Hardware, likewise referenced the shop’s “hardware heritage”—it’s located in an Art Deco–era shophouse, with high, garage-like ceilings and metallic light fixtures. Greg Schuler of Mallorca’s Mistral Coffee noted his design’s embrace of “raw elements”: bare tile floors, exposed ductwork, and plywood shelves.

在2018年出版的学术论文集《全球化布鲁克林：设计全球化城市的美食体验》中，编辑法比奥·帕拉塞科利（Fabio Parasecoli）和马特乌什·哈拉瓦（Mateusz Halawa）提到了“去中心化的千篇一律”。他们写道，2010年代咖啡馆或餐厅的同质化模式“缺乏任何形式的集中协调”，而是以网络动态的方式运作，网络中的每个节点既广播信息，也接收信息。它的美学。这种传播的主要渠道是Instagram，咖啡馆老板们也强调了这一点。顺从的压力是真实存在的。

In a 2018 academic collection titled Global Brooklyn: Designing Food Experiences in Global Cities, the editors Fabio Parasecoli and Mateusz Halawa reference “decentralized sameness.” The homogenous template of 2010s cafés or restaurants “lacks any forms of centralized coordination,” they wrote, but instead operates on network dynamics, with each node in the network broadcasting as well as receiving its aesthetics. The major network through which this distribution takes place is Instagram, a fact that the café proprietors emphasized. The pressure to conform was real.

多伦多极简主义咖啡馆连锁店 Pilot Coffee Roasters 的市场经理 Trevor Walsh 告诉我，过去十年里，Instagram 成了“我们观察全球特色咖啡馆世界的镜头”。“我们希望设计能够与精美的照片相得益彰，营造一个值得分享的氛围。”Pilot 的 Instagram 账号上发布照片，让顾客分享体验，是与其他城市的咖啡馆和咖啡行业同行建立联系的一种方式。但这个平台也带来了跟上潮流的压力。“我们总是迫切地想要创作内容。我们总是觉得我们必须出现在人们的手机里，出现在人们的电脑里，”Walsh 说。他们必须满足算法推送的信息。

Over the past decade, Instagram became “the lens in which we view the global specialty café world,” Trevor Walsh, the marketing manager of Pilot Coffee Roasters, a chain of minimalist cafés in Toronto, told me. “We want to have design choices that play into nice photos, an environment that would be a shareable moment.” Posting photos to Pilot’s Instagram account and having customers share their experiences was a way to connect with cafés and coffee-industry colleagues in other cities. But the platform also created a pressure to keep up. “There’s this constant urgency to be producing content. We are constantly feeling like we have to be in people’s phones, be in people’s desktops,” Walsh said. They had to fill the algorithmic feed.

仅仅作为一家咖啡店存在是不够的；企业必须在互联网上培育平行存在，这是一门完全独立的技能。“感觉上，你必须具备社交媒体敏锐度，必须精通与你的业务相邻但又不直接嵌入业务的领域，才能获得成功和知名度，”沃尔什继续说道。这意味着要达到一些指标，比如在 Instagram 上发布大量带标签的照片，以及在谷歌地图上的企业列表上获得积极的用户评论。这种数字化存在压力的最终结果是形成一种只存在于线上的餐厅形式。“幽灵厨房”是指提供简单外卖食品（如汉堡或披萨）的列表，它们在 Uber Eats 或 DoorDash 上有品牌标识，但没有实际位置；它们由其他餐厅或工业集团厨房经营。这些食品首先以数字内容的形式存在，并通过相同的渠道传播。

Simply existing as a coffee shop isn’t enough; the business has to cultivate a parallel existence on the Internet, which is a separate skill set entirely. “It almost feels like, you must have a social media acumen, you must be savvy in this area that is adjacent to your business, but not directly embedded in your business, in order to be successful and visible,” Walsh continued. That means achieving metrics like plenty of tagged photos on Instagram and positive user reviews on the business’s listing on Google Maps. The end point of this pressure to have a digital presence was a form of restaurant that only existed online. “Ghost kitchens” were listings for simple delivery food, like burgers or pizza, that had brand identities on Uber Eats or DoorDash but no physical location; they were run out of another restaurant or an industrial group kitchen. The food existed as digital content first and traveled through the same channels.

社交媒体敏锐度要求对每个平台的推荐算法和其他任何事情都了如指掌。沃尔什观察到，有些公司可能有很多精彩的故事可以讲述，但他们“并没有尝试跟上这些算法模式，从而让更多受众看到它们”。也许他们发布帖子的频率不够高，或者他们没有跟上 Instagram 等平台推广视频多于静态图片的趋势。这一变化尤其明显，发生在 2022 年左右，当时该平台试图模仿TikTok。即使你自认为知道如何破解算法推广，它也可能失败。正如沃尔什告诉我的那样：“我们投入了大量的时间和精力来创作精彩的内容。但由于算法的缘故，我们发现我们并没有像我们想象的那样吸引到那么多的眼球，有时这可能会让人有点沮丧。” 人气与其说取决于咖啡的味道，不如说取决于它在Instagram照片中的样子。“在线形象”几乎胜过其他一切，这是一条适用于所有行业的法律：酒吧、面包店、时尚精品店，甚至艺术博物馆。（在艺术博物馆中，画廊本身就被设计成用于观看物品的空间，这很有帮助，它们创造了一个完美的视觉框架，可以转化为Instagram。）

Social media acumen requires awareness of each platform’s recommendation algorithm as much as anything else. Walsh observed that some companies may have great stories to tell, but they “are not attempting to keep up with these algorithmic patterns that will allow them to be visible to a larger audience.” Maybe they don’t post often enough, or they don’t keep up with shifts like Instagram promoting videos more than still images, a particularly stark change that occurred around 2022 as the platform attempted to mimic TikTok. Algorithmic promotion can also fall through, even if you think you know how to hack it. As Walsh told me, “We’ve put a lot of time and energy into creating beautiful content. But as a result of that algorithm, we find we’re not necessarily hitting as many eyeballs as we think we could or should, and sometimes that can be a little disheartening.” Popularity has less to do with how the coffee tastes than how it looks in an Instagram photo. Online appearance outweighing almost all else is a law that now applies to all kinds of businesses: bars, bakeries, fashion boutiques, and even art museums. (In the last case, it helps that galleries were already designed as spaces for looking at stuff, creating a perfect visual framing that could be translated to Instagram.)

“我讨厌算法。每个人都讨厌算法，”罗马尼亚布加勒斯特咖啡公司Beans & Dots的老板兼创始人安卡·温古里亚努（Anca Ungureanu）说道。Beans & Dots的前身是一家印刷厂。她的目标是打造“当时在布加勒斯特尚不存在的东西”——一个至少在美学上具有非本土性的空间。它吸引了来自世界各地的人群；当有人在谷歌上搜索布加勒斯特的特色咖啡店时，Beans & Dots就会跳出来。温古里亚努创建了一个Instagram账户，上面全是卡布奇诺快照，积累了超过七千名粉丝。但她感到这个平台剥夺了她通过信息流接触受众的能力，这让她感到沮丧。当她的咖啡馆开始在线销售咖啡时，Facebook和Instagram似乎限制了它们的覆盖范围——除非它们购买广告并提高社交媒体公司自身的利润。这感觉就像算法勒索：要么付钱给我们，要么我们不会推广你。咖啡馆原本用来发展业务、吸引新顾客的工具突然间就变得对它不利了。Facebook 和 Instagram“不让你利用已经建立的社群。从某个时刻开始，事情就变得不公平了，”温古里亚努说道。

“I hate the algorithm. Everyone hates the algorithm,” said Anca Ungureanu, the owner and founder of Beans & Dots, a coffee company in Bucharest, Romania, with its original location in a former printing plant. Her goal was to build “something that did not exist at that moment in Bucharest”—a space that was, at least aesthetically, nonlocal. It draws an international crowd; when someone searches Google for specialty coffee shops in Bucharest, Beans & Dots pops up. Ungureanu developed an Instagram account full of cappuccino snapshots and accrued over seven thousand followers but grew frustrated when she felt that the platform was taking away her ability to access her audience through the feed. When her café started selling coffee online, Facebook and Instagram seemed to throttle their reach—unless they bought ads and boosted the social media company’s own profits. It felt like algorithmic blackmail: pay our toll or we won’t promote you. The tools that had served the café to grow and access new customers were suddenly being turned against it. Facebook and Instagram “don’t let you take advantage of the community you’ve already built. From a certain moment onward, things are unfair,” Ungureanu said.

其他咖啡馆老板也表达了同样的抱怨，比如柏林 Hallesches Haus 的联合创始人吉莉安·梅 (Jillian May)。这家咖啡馆兼精品杂货店于 2014 年开业。店内天花板高耸，空间简朴，拱形窗户环绕，顾客可以买到洒水壶、灯具、陶瓷花盆，以及咖啡或沙拉。这家咖啡馆在 Instagram 上拥有近三万名粉丝。然而，“越来越少”“随着时间的推移，点赞数与我们的用户数量成正比，”梅告诉我。“五年前发布的同一张照片能得到一千个点赞，而今天只能得到一两百个点赞。”她认为这款应用“在强迫用户为转发帖子付费，而我们对此感到不舒服。”对于一个以民主化、用户生成内容为前提的社交网络来说，这种差异就像是一个违背的承诺。我们用户是社交媒体运转的动力，然而，我们也无法完全控制我们在平台上发展的关系，这在很大程度上是因为算法推荐占据了主导地位。

Other café owners made the same complaint, like Jillian May, the cofounder of Hallesches Haus in Berlin, a café and boutique general store that opened in 2014. In its high-ceilinged, austere space set with arched windows, visitors can buy watering cans, lamps, and ceramic planters as well as coffees or salads. It has almost thirty thousand Instagram followers. Yet “there have been fewer and fewer likes over time proportional to our user numbers,” May told me. “The same kind of photo that was posted five years ago would get a thousand likes, while today it receives only one hundred to two hundred likes.” She feels the app is “pushing its users to pay for boosting posts, which we are not comfortable doing.” That discrepancy feels like a broken promise for a social network that was premised on democratized, user-generated content. We users are what makes social media run, and yet we also aren’t given full control over the relationships we develop on the platforms, in large part because algorithmic recommendations are so dominant.

May 观察到了一种或许可以称之为粉丝膨胀的效应。随着平台优先级的改变、一些活跃账户的离开，或者同样的内容技巧不再奏效，高粉丝数量与实际参与度的关联性会随着时间的推移而越来越小。对于过去十年使用过 Instagram 的所有人来说，这种感觉并不陌生。虽然自拍获得的点赞减少可能会损害你的自尊心，但当粉丝的足迹成为企业盈利的途径时，这就会成为一个真正的财务问题，无论是咖啡馆吸引顾客，还是网红销售赞助内容。

May observed an effect that might be called follower inflation. High follower numbers correlate less and less to actual engagement over time, as the platform’s priorities change and as some active accounts leave, or the same content tricks stop working. It’s a familiar feeling for all of us who have been on Instagram over the past decade. While it might hurt your ego to receive fewer likes on a selfie, it’s a real financial problem when that follower footprint is how a business makes money, whether it’s a café attracting visitors or an influencer selling sponsored content.

追求Instagram式的吸引力是一个陷阱：无论是实体空间还是纯数字内容，采用一个可识别的模板所带来的快速增长，最终都让位于每天更新帖子和了解算法最新变化的繁琐工作——需要遵循哪些标签、表情包或格式。数字平台剥夺了企业主的自主权，迫使他们步调一致，而不是追求自己的创意灵感。过于追随潮流也存在风险。陈词滥调并不可取。如果一个比喻过于陈旧，算法受众也不会接受。这就是为什么完美的普通咖啡店设计会不断进行细微调整，增加更多盆栽或减少一些盆栽。在算法推送中，时机就是一切。

Pursuing Instagrammability is a trap: the fast growth that comes with adopting a recognizable template, whether for a physical space or purely digital content, gives way to the daily grind of keeping up posts and figuring out the latest twists of the algorithm—which hashtags, memes, or formats need to be followed. Digital platforms take away agency from the business owners, pressuring them to follow in lockstep rather than pursue their own creative whims. There’s a risk as well in hewing too closely to trends. Cliché is not desirable. If a trope is too stale, the algorithmic audiences won’t engage with it, either. That’s why the perfect generic coffee shop design keeps changing slightly, adding more potted plants or taking a few away. In the algorithmic feed, timing is everything.

另一个策略是保持一致，不去担心潮流或参与度，只坚持自己最了解的——在最深层次上保持个人气质或品牌形象的真实性。京都地下咖啡馆六洋舍就是这样一家店。如果顾客愿意，它完全可以让他们自然而然地融入到它的风格中。它没有Instagram账号。某种程度上，咖啡店同样是物理过滤算法：它们根据人们的喜好对他们进行分类，通过设计和菜单选择悄悄地吸引特定的人群并排斥其他人。从长远来看，这种社区形成可能比获得完美的拿铁艺术和收集 Instagram 粉丝更重要。这就是安卡·温古里亚努在布加勒斯特试图做的事情。“我们是一家咖啡店，在这里你可以遇到和你一样的人，和你有兴趣的人，”她说。她的评论让我想到，一定程度的同质化可能是算法全球化不可避免的结果，仅仅是因为现在有这么多志同道合的人在相同的物理空间中移动，受到相同数字平台的影响。相同性有一种复合的方式。

The other strategy is to remain consistent, not worrying about trends or engagement and simply sticking to what you know best—staying authentic to a personal ethos or brand identity in the deepest sense. Rokuyosha, the basement café in Kyoto, is just such an institution. It can afford to let its customers simply drift to its level, if they so desire. It has no Instagram account. In a way, coffee shops are physical filtering algorithms, too: they sort people based on their preferences, quietly attracting a particular crowd and repelling others by their design and menu choices. That kind of community formation might be more important in the long run than attaining perfect latte art and collecting Instagram followers. That’s what Anca Ungureanu was trying to do in Bucharest. “We are a coffee shop where you can meet people like you, people that have interests like you,” she said. Her comment made me think that a certain amount of homogeneity might be an unavoidable consequence of algorithmic globalization, simply because so many like-minded people are now moving through the same physical spaces, influenced by the same digital platforms. The sameness has a way of compounding.

算法旅游

TOURISM BY ALGORITHM

Yelp 或谷歌地图上的推荐系统通过略微改变人们的行动方向，重塑了城市的地理格局：游客会选择一家咖啡馆而不是另一家，或者即使初来乍到，也能轻松找到自己喜欢的餐厅所在街区。现在，搜索和浏览景点信息就像浏览幻灯片一样简单，游客可以选择去哪里，并将它们作为一种内容形式进行浏览。找到心仪的地点后，他们或许会叫车，顺畅地送达目的地，从而充分利用另一个按需算法市场。这种略微的改变也会改变资金和注意力的流向，使某些地点比其他地点更具优势。如果算法推广持续下去，这种增长将会自我强化，因为注意力会再次引发注意力。

Recommendations on Yelp or Google Maps reshape the geography of a city by slightly redirecting the steps of individuals: a visitor walks to one café instead of another, or easily finds the neighborhood with the kind of restaurants they like even though they are new in town. It’s easier to search and flip through the attractions of a place like a slideshow and choose where to go, perusing them as a form of content. After finding a desirable spot, they perhaps call a car to deliver them there frictionlessly, taking advantage of another on-demand algorithmic marketplace. That slight redirection also shifts the flow of money and attention, privileging some locations over others. If the algorithmic promotion lasts, the growth is self-reinforcing as attention once again begets attention.

Filterworld 提供了一个日益密闭的泡泡，让你可以穿梭于世界之中，你的注意力会顺畅地从一个物体转移到另一个物体，你的肢体动作会通过应用程序引导你前往每个目的地。我亲身经历了旅行中的这种转变。Yelp 的推荐让位于 Airbnb 的短租，这让你更容易感觉自己不像个游客，而更像是任何一个国际化都市某个酷炫街区的临时居民。Uber 逐渐扩展到更多国家，因此我可以使用同样熟悉的应用程序和界面在世界各地叫出租车。随着最近移动支付的普及，我可以用手机支付。在纽约、伦敦或里斯本的商店，甚至地铁上都能买到。（加密货币承诺，尽管并未真正兑现，但它是一种独立于政府的货币，可以在任何地方无缝使用。）这类工具有时会让地点显得毫无意义，因为它们可以通过手机轻松导航。城市成为无所不知屏幕的背景，陷入流动的空间。地理差异让位于数字相似性，当太多人遵循相同的算法路径时，这有时会非常不方便。

Filterworld offers an increasingly airtight bubble in which to traverse the world, your attention smoothly routed from one object to the next, your physical movements guided toward each destination through apps. I’ve experienced this shift in travel myself. Yelp recommendations gave way to Airbnb rentals, which made it easier to feel less like a tourist than like a temporary resident in a cool neighborhood in any global city. Uber gradually expanded into more countries, so I could hail a taxi using the same familiar app and interface across a greater swath of the world. With the recent expansion of mobile payments, I can tap my phone to pay for things at a store or even get on the subway in New York City, London, or Lisbon. (Cryptocurrency promised, although did not really deliver, a government-independent money that could be spent frictionlessly anywhere.) Such tools have a way of making places feel slightly meaningless, since they can be so easily navigated by phone. The city becomes a backdrop to the omniscient screen, falling into the space of flows. Geographical differences give way to digital similarities, which can sometimes be very inconvenient when too many people are following the same algorithmic pathways.

2010年代中期，洛杉矶居民发现地图应用Waze正在破坏原本安静的社区。Waze会根据其他车辆的实时数据自动筛选出理想路线，提供行车路线指引。当高速公路交通拥堵时，该应用会引导驾驶员绕道住宅区街道，以便他们更快地到达目的地。《洛杉矶杂志》 2018年的一篇文章宣称：“算法就是上帝。” 虽然这种技巧在应用人数不多的情况下奏效，但随后也引发了自身的问题。北好莱坞和影城的街区挤满了汽车和卡车，它们经常被困在陡峭的山坡和狭窄的拐角处。事实证明，Waze算法将这些小街道的容量视为与大街道相同；它给出的建议似乎表明这些街道可以再容纳100辆汽车，但实际上它们根本无法承受。交通拥堵造成了噪音污染和事故，以至于当地政府开始与谷歌协商，希望修改数据。附近居民开始自行反击算法，他们登录 Waze 报告虚构的事故和路障，从而向该应用程序发出信号，表示经过他们家的替代路线不会节省任何时间。入侵算法立即改变了他们的实际环境。

In the mid-2010s, Los Angeles residents discovered that the map app Waze, which offered driving directions based on real-time data from other cars automatically filtered into an ideal route, was ruining otherwise quiet neighborhoods. When traffic was bad on the highway, the app would reroute drivers through residential streets in an attempt to get them to their destinations faster. “The algorithm is God,” a 2018 article in Los Angeles Magazine proclaimed. The trick worked while not too many people were doing it, but then it caused its own problems. The neighborhoods of North Hollywood and Studio City became crowded with cars and trucks, which got stuck on steep hills and tight corners. It turned out that the Waze algorithm was treating such smaller streets as having the same capacity as much larger ones; it made recommendations as if the streets could handle the flow of another hundred cars, though of course they could not. The traffic caused noise pollution and accidents, to the point that the local government began negotiating with Google to get the data changed. Neighborhood residents took to fighting back against the algorithm themselves, logging in to Waze to report fictional accidents and roadblocks, thus signaling to the app that an alternative route past their house wouldn’t save any time. Hacking the algorithm made an immediate change in their physical environment.

在这种情况下，吸引算法的注意力会导致数据溢出。机器认为住宅区道路是它需要解决的方程式的便捷解，尽管这个解本身也存在一些问题。（当地居民的幸福感并非算法考虑的变量。）我在伴侣杰西父母位于康涅狄格州韦斯特波特的家附近也观察到了同样的事情。距离……一小时车程在纽约市，许多通勤者会经过韦斯特波特，继续向北或向东行驶。当 I-95 高速公路拥堵时，Waze 或谷歌地图（其母公司于 2013 年收购了 Waze 并开始整合其数据）会引导司机穿过这座拥有两百年历史的古镇中心，走上原本只供本地车辆通行的道路。双车道桥梁会堵塞，较短的红绿灯勉强让几辆车通过，然后再次变成红灯。最终，时间并没有被节省，市中心的生活质量也受到了影响。

In this case, attracting the algorithm’s attention caused an overflow. The machine saw the residential roads as convenient solutions for the equation it had to solve, though the solution had its own discontents. (Local happiness was not a variable the algorithm considered.) I observed the same thing happening around my partner Jess’s parents’ home in Westport, Connecticut. An hour outside of New York City, many commuters drive past Westport on their way farther north or east. When the I-95 highway is crowded, Waze or Google Maps (whose parent company acquired Waze and began incorporating its data in 2013) sends the drivers through the quaint center of the two-hundred-year-old town, on roads meant only for local traffic. The two-lane bridge becomes blocked, and the short traffic lights barely let through a few vehicles before turning red again. In the end, no time is saved, and the quality of life downtown suffers.

正是因为应用程序本身，这种奇特的运动才能如此迅速而激烈地发生，这又是一种“去中心化的千篇一律”，但它的本质在于行动而非审美。Waze 被动地引导司机，他们甚至可能没有意识到前方即将出现减速，选择它认为最佳的替代路线。这款普遍可靠的应用程序的存在取代了我们对决策及其所需的判断。（虽然高速公路的选择并非真正的文化差异问题，但它体现了算法推荐取代个人品味的趋势。）我严重依赖谷歌地图来指引我到达目的地，因为我早已忘记了曾经记住的任何行车路线。查阅高速公路地图已是童年时代的遥远记忆；我已经完全臣服于自动导航。Waze 算法是资本和数据的新聚合体，它以意想不到的方式扭曲着周围的世界。它既能通过无处不在的个人智能手机生成实时交通数据，又能根据交通模式、油耗和通行费等因素，制定驾驶建议。然而，驾驶员对这些变量及其权重的计算方式大多一无所知；他们只能感受到最终结果。

It’s only because of the app itself that this strange movement can happen with such speed and intensity, another kind of “decentralized sameness,” but of action rather than aesthetic. Waze passively directs drivers, who may not even be aware of a looming slowdown ahead, to take the alternate route that it determines to be optimal. The presence of the generally reliable app replaces our need for decision-making and the judgment that it requires. (While highway choice is not really a matter of cultural distinction, it’s an echo of algorithmic recommendation’s supplanting of personal taste.) I rely heavily on Google Maps to guide me to my destination, having long since forgotten any driving directions I once memorized. Consulting maps of highways is a distant memory from childhood; I have fully surrendered to automated navigation. The Waze algorithm is a new agglomeration of capital and data that warps the world around it in unexpected ways. It both generates the real-time traffic data, through the omnipresence of personal smartphones, and leverages it to create driving recommendations, calculated by traffic patterns, gas efficiency, and tolls. Yet the driver is largely unaware of these variables and how they are being weighed; they only experience the end result.

整个冰岛都经历了类似 Waze 高速公路效应的现象，尽管这是他们刻意营造的。这是一个天然的 Instagram 陷阱。这座岛屿和地球上任何地方一样，不太可能成为吸引人类的目的地。它独自漂浮在大西洋北部，由活火山、险峻的冰川和险峻的峡湾组成。事实上，直到公元九世纪左右，这里都完全无人居住。一位名叫 Garðar Svavarsson 的瑞典维京人是第一个在岛上航行的人。公元 870 年，人们开始环岛开辟道路，最早的定居点雷克雅未克也于 874 年出现。（该岛名字的意思是“烟雾湾”，之所以这样称呼是因为酋长 Ingólfr Arnarson 在登陆之前注意到了地热喷口冒出的蒸汽。）1703 年首次正式测量岛上人口时，只有五万多一点。即使在今天，人口也不到四十万，是世界上最低的人口密度之一，每平方英里只有八人左右。然而，冰岛每年仍吸引着超过 200 万游客（是当地人的五倍），每个游客至少在冰岛住一晚。2019 年，旅游业占该国出口收入的 35%，在新冠疫情导致国际旅行中断之前达到顶峰。旅游业比任何其他行业都大，这意味着冰岛最赚钱的产品就是自我推销，提供曾经人迹罕至的地方的体验。

The entire nation of Iceland experienced something like the Waze highway effect, albeit a version that it intentionally cultivated. It’s an organic Instagram trap. The island is as unlikely a destination to attract humans as just about anywhere on the planet. Floating by itself in the northern stretch of the Atlantic Ocean, it’s made up of active volcanoes, forbidding glaciers, and precipitous fjords. In fact, until around the ninth century CE, it was entirely uninhabited. A Swedish Viking named Garðar Svavarsson was the first to navigate around the entire island in 870, and the earliest settlements in Reykjavík followed shortly after in 874. (Its name translates to “cove of smoke,” so called because the chieftain Ingólfr Arnarson noticed steam from geothermal vents before landing at the site.) When the island’s population was first officially measured in 1703, there were a little over fifty thousand residents. Even today the population numbers under four hundred thousand, with one of the lowest densities in the world at around eight people per square mile. And yet Iceland has regularly attracted over two million tourists per year (five times the number of locals), who stay at least one night in the country. Tourism accounted for 35 percent of the nation’s export revenue in 2019, reaching a peak just before the COVID-19 pandemic halted international travel. It was larger than any other industry, meaning that Iceland’s most profitable product was selling itself, offering experiences of a once-inaccessible place.

旅游业并非冰岛一直以来的支柱产业。游客数量直到2000年左右才超过当地人口，而真正实现增长是在2010年左右。从那一年开始，旅游业呈指数级增长，就像一个模因的传播一样。讽刺的是，一场灾难却开启了21世纪的旅游业热潮。第一个原因是2010年3月的埃亚菲亚德拉冰盖火山喷发，由于弥漫在空气中的火山灰会损坏飞机引擎，导致4月份欧洲各地航班停飞一周。（就冰岛本身而言，火山喷发发生在一个非常偏僻的农业社区，只需要疏散大约八百人——没什么大不了的。）如此引人注目的新闻事件在世界各地报道，所有电视新闻都刊登了冰岛的照片，而冰岛通常不会受到如此全球媒体的关注。突然之间，数百万人看到了这片原始的景观：远处隐约可见的冰川、奔腾的瀑布和天然的温泉。“因为火山爆发，全世界的目光一下子都集中在我们身上，”雷克雅未克市旅游创新部门的卡伦·玛丽亚·约斯多蒂尔告诉我。

Tourism wasn’t always such a cornerstone industry for Iceland. Travelers only exceeded the local population around the year 2000, and growth really took off around 2010. From that year, the chart goes exponential, like the spreading of a meme. Ironically, it was a disaster that kicked off the twenty-first-century tourism boom. The first cause was the eruption of the volcano Eyjafjallajökull in March 2010, which grounded flights across Europe for a week in April, since the volcanic ash filling the air would damage airplane engines. (As far as Iceland itself was concerned, the eruption happened in a very isolated farming community and required the evacuation of only around eight hundred people—no big deal.) Such a dramatic news event was covered across the world, and all those TV news spots ran photos of Iceland, which wasn’t usually the subject of such global media exposure. Suddenly, many millions of people were seeing the pristine landscape, the glaciers looming in the distance, the torrential waterfalls, the natural hot springs. “Because of the volcano eruption, the whole world suddenly had its eyes on us,” Karen María Jónsdóttir, who works in tourism innovation for the city of Reykjavík, told me.

一些冰岛人认为，旅游业繁荣的另一个原因是Instagram。火山爆发后的几年里，数百万人试图重现那些田园诗般的冰岛风光。为自己的账号上传风景照片。将度假照片发布到网上，是对上世纪70年代典型的家庭幻灯片放映的21世纪升级。看到一张精彩的旅行照片，尤其是在网络社交平台上，会引发一种发自内心的“错失恐惧症”（FOMO），即害怕错过。你想知道照片是在哪里拍摄的，拍摄者是如何到达那里的，以及他们在那里住在哪里。多亏了Instagram的标签和位置标签以及相应的落地页，点击链接就能轻松查看特定地点最受欢迎的算法推荐照片。当然，冰岛的美丽并非源于Instagram，就像咖啡馆的设计是为了迎合Instagram一样。它的美丽源于各种地质和地理因素。但Instagram精心设计并突出了冰岛的自然资源，该平台的推荐将冰岛热门景点的标志性图像推到了数亿用户的信息流顶部，使这些图像成为了冰岛事实上的代表。

The other cause of the tourism boom, according to some Icelanders, was Instagram. Over the years following the eruption, millions of people would attempt to re-create those idyllic Icelandic landscape photos for their own accounts. Posting your vacation photos online was the twenty-first-century update to the archetypal 1970s at-home slideshow. Seeing a great travel photo, especially in a social space online, induces a visceral sense of FOMO, the fear of missing out. You want to know where it was taken, how the person got there, and where they stayed while they were there. And thanks to Instagram’s hashtags and location tags, with their attendant landing pages, it was very easy to tap a link and see the most popular algorithmically recommended photos from a given location. Of course, Iceland wasn’t beautiful because of Instagram, the way cafés were intentionally designed to be Instagram-ready. It was beautiful because of various accidents of geology and geography. But Instagram framed and highlighted its natural assets, and the platform’s recommendations propelled iconic images of Icelandic hot spots to the top of the feed for hundreds of millions of users, turning those images into the de facto representations of the country.

就像Galaxie 500乐队的达蒙·克鲁科夫斯基（Damon Krukowski）发现，与其说乐队成员自己选择的代表自己的单曲，不如说他们乐队的音乐更受Spotify算法驱动的热门歌曲的认同。Instagram动态也开始为冰岛塑造新的公众形象。在冰岛航空工作了几十年的迈克尔·劳切森（Michael Raucheisen）向我解释说，社交媒体开始推动冰岛航空对游客需求以及如何吸引游客的认知。“我们从乘客身上看到的视觉效果几乎比我们库存的照片更令人惊叹，”他说。该航空公司开始在其机上杂志上发布来自Instagram的照片。“世界各地的人们都可以在社交媒体上分享冰岛的美丽，这让我们的工作变得轻松多了，”劳切森说。同样，Instagram并非创造了旅游业，但它为旅游业带来了巨大的机遇。冰岛航空提供廉价航班和中途停留优惠，旨在帮助游客顺道游览。雷克雅未克周边新建了现代化的酒店。为游客专门设计了主题酒吧，例如“雷克雅未克美国酒吧”。

The same way that Damon Krukowski of Galaxie 500 found his band’s music identified more with its algorithmically propelled top song on Spotify than the singles the musicians had chosen to represent themselves, the Instagram feed began to create a new public identity for Iceland. Michael Raucheisen, who has worked for the airline Icelandair for decades, explained to me that social media began driving Icelandair’s perception of what tourists were looking for and how to attract them. “The visuals that we saw from passengers were almost more astounding than the photos we had in our stock,” he said. The airline began publishing photos sourced from Instagram in its in-flight magazine. “The fact that people all over the world can share the beauty of Iceland on their social media channels makes our job a lot easier,” Raucheisen said. Again, it wasn’t that Instagram created the tourism, but it gave the tourism industry an enormous opportunity. Icelandair offered cheap flights and layover deals designed to help travelers stop by. New, modernized hotels were built around Reykjavík. Bars popped up themed for the visitors, with names like American Bar Reykjavík.

社交媒体也吸引了新的游客群体，他们的行为举止出人意料。在繁荣之前的几十年里，大多数前往冰岛的游客也来自邻近的北欧国家。例如德国和法国。旅行者往往会停留很长时间，计划在该国偏远地区进行探险。最近，主要的旅游群体变成了美国人、英国人和中国人。他们停留的时间较短——有些人是被中途停留的优惠吸引，在前往其他地方（通常是欧洲大陆）的途中停留——而且他们更倾向于选择雷克雅未克。从城市酒店出发的巴士之旅会将旅游团送到主要景点、间歇泉和瀑布，然后再将他们安全地送回房间。算法推荐也在塑造这些行程。在线旅行社（OTA），如Kayak和Booking.com，在2010年代不断发展，每年处理数亿笔旅行预订。这些网站集搜索引擎、用户生成的评论网站和推荐系统于一体，为航班、酒店和城市提供五星平均评分，就好像它们是Netflix上众多可供选择的节目一样。通过这些渠道引导的游客越多，他们在冰岛的路线就越趋同，就像 Waze 引导汽车穿过韦斯特波特市中心一样。一个优化版的冰岛消费模式应运而生。

Social media attracted new demographics of tourists, too, who behaved in unexpected ways. In the decades before the boom, most tourists to Iceland were from neighboring Nordic countries as well as Germany and France. The travelers tended to stay for long periods of time, planning expeditions in far-flung parts of the country. More recently, the major groups have changed to American, British, and Chinese. They stay for shorter lengths of time—some drawn in by the layover offers, stopping on their way to somewhere else, often mainland Europe—and they stick closer to Reykjavík. Bus trips leaving from city hotels send tourist groups out to the major sights, the geysers, and waterfalls, before depositing them safely back at their rooms. Algorithmic recommendations were shaping those itineraries, too. Online travel agencies, or OTAs, like Kayak and Booking.com, were growing over the 2010s, accounting for hundreds of millions of travel bookings per year. The websites were a combination of search engines, user-generated review websites, and recommendation systems, giving flights, hotels, and cities average ratings out of five stars as if they were so many Netflix shows to choose from. The more tourists who were routed through these channels, the more their pathways through the nation became homogenized, like Waze routing cars through downtown Westport. An optimized version of Iceland emerged for consumption.

就像那些“普通咖啡店”的顾客一样，算法驱动的全球旅行者也享有特权。想要在世界各地自由出行，需要持有合适的护照（通常是美国护照），以及符合性别和肤色的表达方式。举个例子，黑人用户长期以来一直难以使用 Airbnb，因为那里的房东带有种族歧视的倾向，会拒绝他们的预订请求，直到该平台移除了显示用户面部信息的个人资料。并非每个人都能轻易被视为普通人，或融入看似毫无摩擦的空间。为某一类用户提供便利并不意味着为所有用户提供便利。

Much like the denizens of Generic Coffee Shops, the algorithmic global travelers were a privileged set. Being able to move freely across the world necessitated the right passport (likely an American one) as well as the right gender expression and skin color. Black users, to name just one example, long had difficulty using Airbnb because of racist hosts, who would deny their booking requests, until the platform eliminated profiles that displayed users’ faces. Not everyone is able to easily be perceived as generic or fit into seemingly frictionless spaces. Convenience for one group of users doesn’t mean convenience for all of them.

Booking.com 将“黄金圈全日游”列为雷克雅未克“最受欢迎”景点榜首，这条线路涵盖了冰岛最著名的景点，包括豪卡达鲁尔间歇泉、黄金瀑布和辛格韦德利国家公园——公元 930 年第一届冰岛议会的所在地。96% 的游客喜欢这条线路，网站在搜索结果下方的醒目位置展示了这条线路。（相比之下，只有 69% 的游客喜欢列出的三小时观鲸之旅——这暗示着，别费心了。）OTA如何筛选这些产品，对游客的选择有着巨大的影响，从而也决定了哪些企业（例如酒店和旅行社）能够生存下去。当我采访冰岛旅游局局长斯卡菲丁·伯格·斯泰纳森（Skarphéðinn Berg Steinarsson）时，他很快就对这些服务提出了批评。“他们的经济回报来自快速销售和简单的产品。他们无休止地推广十大榜单，因为那只是简单的产品，”斯泰纳森说道。其动机并非提供特别独特的体验，而是为了说服尽可能多的网站访问者点击“购买”按钮，参与到内容中。

At the top of its list of “most popular” attractions in Reykjavík, Booking.com puts a “Full-Day Tour of the Golden Circle,” a circuit of Iceland’s most famous sights, including the Haukadalur geysers, the Gullfoss waterfall, and Þingvellir National Park, the site of the first Icelandic parliament in 930 CE. The tour was liked by 96 percent of visitors, as the site displays prominently under the search results. (By contrast, only 69 percent liked the listed three-hour whale-watching tour—the implicit message being, don’t bother.) How the OTAs sort these offerings have an outsize impact on which ones tourists select, and thus which businesses, like hotels and tour operators, survive. When I spoke to Skarphéðinn Berg Steinarsson, the director general of the Icelandic Tourist Board, he was quick to criticize the services. “Their financial reward comes from quick sales and simple products. They are endlessly promoting top ten lists, because that’s a simple product,” Steinarsson said. The incentive isn’t to offer a particularly unique experience; it’s to convince the most website visitors possible to click the Buy button, to engage with the content.

这种推荐信息流肤浅，只会引发快速、被动的决策；作为互联网用户，我们被灌输的观念是，最靠前的结果才是最好的。但我们却忽略了深度，忽略了那些鲜为人知的点。2006年，《连线》杂志主编克里斯·安德森出版了他的畅销书《长尾理论》，他认为互联网的广度使得利基企业、产品和内容得以蓬勃发展。流行度是一条曲线，向图表左侧呈指数上升；安德森预测，曲线较长且平坦的部分，包含各种各样但相对冷门的事物，将能够以一种新的方式持续发展，因为消费者总能在网上找到他们正在寻找的特定内容。安德森甚至在21世纪初就观察到了算法推荐的自我延续效应，它推动了1985年出版的一本原本默默无闻的书籍《触及巅峰》的销量，最终登上了畅销书排行榜。“大众市场正在变成利基市场的集合，”安德森写道。

Such feeds of suggested information are superficial and inspire quick, passive decision-making; as Internet users, we have been taught that the top results are the best ones. But we are missing the depth, the lesser-known spots. In 2006, the Wired editor-in-chief Chris Anderson published his hit book The Long Tail, arguing that the breadth of the Internet made it possible for niche businesses, products, and content to thrive. Popularity was a curve rising exponentially toward the left side of a graph; Anderson predicted that the long, flat part of the curve, with diverse but relatively unpopular things, would be able to sustain itself in a new way, because consumers would always be able to find the specific thing that they were seeking online. Anderson observed the self-perpetuating effects of algorithmic recommendations even on Amazon in the early 2000s, as they drove sales of a previously obscure book from 1985, Touching the Void, which eventually hit bestseller lists. “The mass market is turning into a mass of niches,” Anderson wrote.

在“过滤世界”时代，这种效应在某些方面得到了证实。TikTok 确实催生了小众内容制作的职业发展；我喜欢关注的一位创作者就靠制作北极岛屿日常生活的视频维持生计。但如今无处不在的算法推荐也让长尾图的左侧变得更大，因为它们甚至将不感兴趣的消费者引导到已被证明受欢迎的标准化内容集。来自视频游戏流媒体平台 Twitch 的泄露数据显示，只有收入最高的 0.01% 的 Twitch 创作者能够赚取美国中位数收入。安德森可能没有想到，关注度与利润会如此紧密地联系在一起，这得益于广告，而不是书籍或DVD销售。“大众文化不会衰落，只是大众性会下降，”安德森预测道。我们现在知道事实并非如此。如果说有什么不同的话，那就是大众文化近年来在审美上比以往任何时候都更加同质化。

In the Filterworld era, this effect has proven true in some ways. TikTok certainly enables a career in niche content production; one creator I enjoy following sustains herself by making videos about daily life on an Arctic island. But now-ubiquitous algorithmic recommendations have also made the left side of the long-tail graph that much larger, as they route even uninterested consumers toward the standardized set of content that has already proven popular. Leaked data from the video-game streaming platform Twitch suggested that only the top .01 percent of Twitch creators were able to earn the median U.S. income. Anderson may not have considered that attention would equate so closely with profit, driven by advertising rather than, say, book or DVD sales. “Mass culture will not fall, it will simply get less mass,” Anderson predicted. We know now this is not the case. If anything, mass culture lately appears more aesthetically homogenous than ever.

斯坦纳森对冰岛旅游业在OTA推荐浪潮中呈现的长尾效应感到惋惜。“太多地方不在那些名单上，也永远不会出现在那些名单上，”他说。“你需要挖掘多深才能发现真正值得一去的地方？”游客的体验被扁平化，并非因为他们别无选择，而是因为数字平台让跟随其他人的脚步变得异常便捷，这就像老式旅游指南的涡轮增压版，更具强制性。OTA和Instagram都助长了冰岛成为“过度旅游”的象征。“过度旅游”一词由旅游媒体企业家拉法特·阿里于2016年提出，如今越来越多地被用来描述那些因大量游客涌入而遭到破坏或不可逆转地改变的地方。数字内容有望实现可扩展性：只要服务器空间足够，同一个文件就可以服务于无数消费者。更大的受众群体不会改变体验。但实体场所并非可扩展的。

Steinarsson mourned for the long tail of Icelandic tourism against the tide of OTA recommendations. “There are so many places that are not on those lists and never will be on those lists,” he said. “How deep do you have to dig before you start seeing places that would really be something special?” The tourists’ experiences are flattened, not because they have no other option but because the digital platform has made it incredibly convenient to simply follow in the steps of everyone else, a turbocharged, more coercive version of the old-school tourist guidebook. Both OTAs and Instagram have contributed to turning Iceland into a symbol of “overtourism,” the term coined by travel-media entrepreneur Rafat Ali in 2016 and increasingly used to describe places that are being damaged or irrevocably altered by the sheer volume of travelers moving through them. Digital content promises to be scalable: the same file can serve an infinite number of consumers, as long as the server space exists to host it. Larger audiences don’t change the experience. But physical places are not scalable.

2019年，为了完成一个报道任务，我前往冰岛体验那里的旅游业繁荣。疫情导致全球旅行全面暂停之前，冰岛的旅游业正达到顶峰。我没有选择酒店，而是选择住在爱彼迎 (Airbnb)，在雷克雅未克市中心的一小块地段选择了一套公寓。通过爱彼迎的搜索功能，我确认公寓靠近一家名为“雷克雅未克烘焙店”(Reykjavík Roasters) 的工业风咖啡店，这家咖啡店是我事先在谷歌地图上找到的。爱彼迎的公寓以工业风阁楼为原型，落地窗外是微型天际线的景色。在装饰方面，这间毫无特色的公寓挂着一幅巨大的布鲁克林大桥照片，让我感觉离开纽约后哪儿也没去。尽管我选择了这种审美，但对于世界上地理位置较为偏僻的地方之一来说，这种标准化似乎有些过头了。

In 2019, on a story assignment, I traveled to Iceland to experience its tourism boom, which was hitting a peak just prior to the pandemic halting global travel altogether. I stayed in an Airbnb rather than a hotel, selecting an apartment in the small stretch of downtown Reykjavík. Through the Airbnb search function I made sure the apartment was close to an industrial-chic coffee shop called Reykjavík Roasters, which I had identified in advance on Google Maps. The Airbnb was modeled on an industrial loft, with floor-to-ceiling windows opening to a view out over the miniature skyline. For decor, the identity-less apartment had a huge print of a photograph of the Brooklyn Bridge, which made me feel like I hadn’t gone anywhere after leaving New York. Even though I had selected for that kind of aesthetic, the standardization seemed excessive for one of the more physically isolated places in the world.

我从雷克雅未克出发的第一次探险是乘坐巴士，在黄金圈进行为期一天的旅行。感觉就像回到了小学时代——学校实地考察时，我们这群游客满怀期待地坐在座位上，一位表情极其淡定的导游一遍遍地向我们介绍目的地。巨大的黄金瀑布（Gullfoss）气势磅礴，每秒有110立方米的水从地缝中流过，令人叹为观止。即使没有背景——覆盖着翠绿苔藓的岩石峭壁一直延伸到田野——仅凭这巨大的水量和如同音乐会扬声器回声般震耳欲聋的轰鸣声，就已经足够震撼人心。但我们团里的大多数人都用手机相机欣赏着这番景象。他们捕捉到的景色，甚至连角度都和瀑布的Instagram页面上反复出现的一模一样。他们进一步复制这张照片，确保它作为冰岛标志性景观的地位。

My first expedition out of Reykjavík was the daylong bus trip around the Golden Circle. It felt like a throwback to elementary-school field trips, the group of us tourists sitting in our seats expectantly as an extremely stoic tour guide repeated the basic facts of where we were headed. Gullfoss, the enormous waterfall, was an awe-inspiring sight, a crack in the earth that 110 cubic meters of water flow through per second. The sheer volume of water and the roar that filled the air like speaker feedback at a concert would have been impressive enough even without the setting: rock crags covered in verdant green moss extending into fields. But much of my group was looking at the view through their phone cameras. The vista that they were capturing, down to the angle, was the same one that appears over and over again on the falls’ Instagram page. They were further replicating the image, ensuring its dominance as a generic symbol of Iceland.

这让我想起了唐·德里罗 1985 年的小说《白噪音》，书中的主人公是一位大学教授，他和同事默里一起到乡下参观“美国被拍照最多的谷仓”。这个谷仓除了它的恶名之外没有什么特别之处——一个虚构的前互联网模因。默里观察着谷仓周围的摄影师人群，说道：“我们来这里不是为了捕捉图像，而是为了维护图像。每一张照片都强化了这种氛围。”“没有人看到谷仓，”他总结道。“他们在拍摄拍照的照片。”在 Filterworld 中，很难将某事物的本质或其现实与其在关注度方面的受欢迎程度区分开来。仅仅是受欢迎程度常常被误认为是意义或重要性，就像德里罗的谷仓一样。

It reminded me of Don DeLillo’s 1985 novel White Noise, in which the college-professor protagonist travels into the countryside with his colleague Murray to see “The Most Photographed Barn in America.” Nothing makes the barn particularly remarkable except its notoriety—a fictional pre-Internet meme. Observing the crowd of photographers around the barn, Murray says, “We’re not here to capture an image, we’re here to maintain one. Every photograph reinforces the aura.” “No one sees the barn,” he concludes. “They are taking pictures of taking pictures.” In Filterworld, it becomes hard to separate the nature of something, or its reality, from its popularity in terms of attention. Popularity alone often gets confused for meaning or significance, as in the case of DeLillo’s barn.

黄金瀑布既要保护自己免受摄影师的侵害，也要保护摄影师免受他们自己的侵害。瀑布旁的小路沿途竖起了围栏，防止游客靠近边缘，也防止踩到当地的苔藓，这些苔藓可能需要一个世纪才能重新长出来。黄金瀑布周围发生过致命事故，这与美国徒步旅行者自拍时坠崖身亡的事件如出一辙。“如果你摔下去，就不可能找到你。你就消失了，”旅游局局长斯坦纳松说。“那张照片对你来说意味着什么？”他问我。唯一的逻辑是，照片本身比景色本身更重要。

Gullfoss has had to defend itself from photographers—as well as defend the photographers from themselves. Fences have popped up along the path running past the waterfall to prevent tourists from getting anywhere close to the edge or stepping onto the local moss, which can take a century to grow back. There have been fatal accidents around Gullfoss, echoes of American hikers falling off cliffs to their deaths while taking selfies. “If you fall over, it’s impossible to find you. You’re just gone,” Steinarsson, the head of the tourism board, said. “What is that picture going to do for you?” he asked me. The only logic is if the photograph matters more than the sight itself.

抵达冰岛的游客数量一直在缓慢增长毁掉了游客们最初来此观赏的景象：岛上壮丽的自然风光。在一些景点，比如豪卡达鲁尔间歇泉，人们无休止的脚步已经把地面变成了泥泞。在夏季淡季，地面通常有机会干涸，但旅游业越来越趋于全年开放，草坪根本来不及恢复。“如果那里人流太多，它就会被彻底毁掉，”斯坦纳森说。

The sheer volume of tourists arriving in Iceland has been slowly ruining what the tourists came to see in the first place: the island’s natural grandeur. At some sites like the Haukadalur geysers, the relentless plodding of feet has turned the ground into mud. In the summer off-season, it usually has a chance to dry out, but tourism is increasingly year-round and the turf doesn’t have time to recover. “If there’s too much traffic there it would simply spoil,” Steinarsson said.

数字平台驱动的旅游业不仅造成了字面意义上的扁平化，也带来了无数双脚和车轮的物理损伤，以及一种隐喻意义上的扁平化。正如我在其他文化形式中指出的那样，网络的无处不在会使其主体变质，将其精髓磨蚀成更容易被消费的东西，或者干脆导致其被消费直至消失。在“过滤世界”（Filterworld）中，独特性是一种转瞬即逝的产品，它会被算法推荐所抓住，被越来越多地推荐，直到它的独特之处被毁掉或沦为陈词滥调。例如，以白色建筑和蓝色穹顶而闻名的希腊圣托里尼岛，在2019年不得不张贴标语，要求来访的Instagram用户停止擅自闯入屋顶拍摄完美照片。屋顶不仅仅是数字内容；它们是当地居民的家园。 2023年，在疫情过后旅游业复苏之际，圣托里尼岛发布了一段TikTok视频，视频中，这种扁平化现象愈演愈烈，屋顶上排起了长队。视频发布者妮基·吉布森（Nikki Gibson）在画外音中说道：“仅仅因为景色优美而去观赏，这并非旅游的真正意义所在。”

Tourism driven by digital platforms causes a literal flattening, in the physical damage of so many feet and vehicle wheels, as well as a metaphorical flattening. As I have pointed out with other forms of culture, online ubiquity has a way of denaturing its subject, wearing away its essence into something more easily consumable, or else simply causing it to be consumed until it is gone. In Filterworld, uniqueness is an ephemeral product that gets seized on by algorithmic recommendations, recommended more and more until whatever was unique or special about it is either ruined or becomes cliché. The Greek island of Santorini, for example, famous for its white buildings and blue domes, had to post signs in 2019 demanding that visiting Instagrammers stop trespassing on rooftops to capture the perfect photo. The rooftops weren’t just digital content; they were local residents’ homes. A TikTok video from Santorini in 2023, during a post-pandemic resurgence of tourism, observed that this flattening had only intensified, with crowds lining up on the rooftops. “Just going somewhere to visually consume it because it’s beautiful is not really what tourism should be about,” the poster, Nikki Gibson, said in voiceover.

这种Instagram化的过程，在我去过的两处冰岛温泉中都体现得淋漓尽致。第一处是蓝湖温泉，是著名的自拍圣地。它位置便利，靠近机场，周围环绕着一片宽阔的温泉池，周围是一排风景如画的玻璃建筑，池水环绕在黑暗、参差不齐的火山岩中。淡蓝色的池水中升起的蒸汽，在异域风情的映衬下，构成了完美的拍照背景。虽然蓝湖温泉并非纯天然温泉——它是由地热发电厂的径流加热的——但我在那里度过了一段非常愉快的时光，在热水中嬉戏，游到酒吧，用我买的防潮项链，戴着塑料套的手机拍照。我往脸上涂抹了一层我把冰岛的白色粘土涂成白色，然后自拍了一张，发到了Instagram上。我周围的游客也都在做同样的事情。在蓝湖，你永远不会忘记自己是一名游客，这里每年接待约一百万游客。这本身没什么不好，但对于那些以追求更“真实”体验、试图融入当地人而自豪的旅行者来说，这可能会感觉很奇怪。（不过，这确实有助于缓解时差反应。）

This process of Instagrammification was visible in two Icelandic hot springs that I visited. The first, the Blue Lagoon, is an established site for selfies. Conveniently located near the airport, it’s a set of picturesque glassy buildings around a sprawling pool set amidst dark, jagged volcanic rock. The steam rising from the light-blue water against the alien landscape makes for a perfect photo backdrop. Though the Blue Lagoon is not a fully natural hot spring—it’s heated by the runoff of a geothermal power plant—I had a perfectly pleasant time there, bobbing in the hot water, swimming up to the bar, and taking photos with my phone in a plastic case attached to a necklace that I bought to keep it dry. I daubed my face with whitish Icelandic clay and took a selfie that I posted on Instagram. All around me visitors were doing the same thing. You never forget you’re a tourist at Blue Lagoon, which sees around a million guests annually. It’s not a bad thing, but it can feel strange for travelers who pride themselves on seeking out more “authentic” experiences in which they try to be surrounded by locals. (Though it is good at curing jet lag.)

我参观的另一个温泉在我的长途巴士之旅中被称为“秘密泻湖”，尽管它的当地名称是Gamla Laugin，或者简称为“古老的游泳池”。事实上，它是冰岛最古老的温泉；这个由间歇泉自然注水的游泳池始建于1891年，每24小时更换一次水。1909年，这里举办了冰岛第一堂游泳课。在废弃数十年后，它于2014年重新开放，当时只有一栋单层建筑，内设储物柜和淋浴间。与蓝湖不同，它的设计并非为了拍照。水是深蓝色的，而不是浅蓝色的，周围环绕着平坦的石床和长满青苔的小丘，一条小小的木板路穿过潺潺的间歇泉。游泳池历史上曾使用过一座没有门窗的破旧石棚，赋予它一种古老而永恒的感觉。它很有当地特色。我有点尴尬，我们这个由十几个外国人组成的旅行团，在工作日的下午抵达，似乎打扰了几个年纪较大的独自游泳者的日常活动。但我享受着这里宁静的氛围，以及除了欣赏之外，这里不需要我做任何其他事情的感觉。我潜入水中，游到周围有足够的空间。没有人挥舞着手机。

The other hot spring I visited was billed on my extended bus tour as the “Secret Lagoon,” though its local name is Gamla Laugin, or simply “the old swimming pool.” In fact, it is the oldest in Iceland; the pool, fed naturally by geysers that replace its water every twenty-four hours, was first built in 1891. In 1909, it hosted Iceland’s first swimming lessons. After many decades of disuse, it reopened in 2014 with one single-story building housing lockers and showers. Unlike the Blue Lagoon, it was not designed to be photographed. The water was dark, not light blue, and it was surrounded by a bed of flat stones and mossy hillocks, with a tiny boardwalk running through the burbling geysers. A decrepit stone shed without doors or windows remained from the pool’s historical use, giving it a sense of age and permanence. It felt local. I was a little embarrassed that our tour group of a dozen foreigners, arriving on a weekday midafternoon, seemed to be disrupting the routines of a handful of older solo swimmers. But I enjoyed the peaceful atmosphere and the sense that the place didn’t demand anything from me other than appreciating it. I sank into the water and swam to the edge with plenty of space around me. No one was waving a phone around.

虽然我很高兴自己到达了那里，但我也感到一种迫在眉睫的威胁：像我这样的游客蜂拥而至，会削弱老劳金的独特性，把它变成另一个蓝湖，一个拍摄TikTok视频、吸引Instagram评论的地方。Instagram上关于旅行的精彩照片，最常见的评论总是“这是哪里？”，这又暗示着另一个问题：我怎么才能去那里？旅行总是伴随着某种矛盾心理：知道自己是游客，但又渴望与不同的地方直接接触。然而，算法推荐已经自动化了口口相传的过程，将旅游业变成了一条传送带，更深地渗透到当地的文化生态系统中——让地方，或者至少……至少他们的经历，在过程上更相似。这个世界只是内容的潜在素材，吸引一些点赞，为数字平台赚钱，然后消失，不太可能再被看到。为了暂时摆脱这种循环，我离开了Gamla Laugin，再也没有发布任何关于它的内容。

Though I was happy I made it there, I also felt the looming threat of a flood of tourists like me detracting from Gamla Laugin’s uniqueness, turning it into another Blue Lagoon, a place to shoot TikTok videos and attract Instagram comments. The most common comment on a dramatic travel Instagram shot is always “Where is this?,” implying another question: How can I get there, too? A certain ambivalence has always been a part of travel, knowing you’re a tourist but also craving direct contact with a different place. Yet algorithmic recommendations have automated that process of word of mouth and turned tourism into a conveyor belt that intrudes much deeper into local cultural ecosystems—making places, or at least the experience of them, more similar in the process. The world is just potential fodder for content that attracts a few likes, makes money for digital platforms, and then evanesces, unlikely to ever be seen again. Wanting for a moment to avoid that cycle, I left Gamla Laugin and never posted anything about it.

AIRBN 迁移

AIRBNB MIGRATION

在新冠疫情期间，塑造旅游业的算法力量也迅速开始影响美国国内的人口流动。短短几周内，许多人评估房地产的方式就发生了翻天覆地的变化。人口密集的城市曾一度成为越来越理想的居住地，但由于医院人满为患，以及隔离期间的幽闭恐惧症，人口密度的吸引力大大降低。根据美国人口普查局的估计，2020年和2021年，人​​口明显向郊区和农村地区转移。俄勒冈州本德市、缅因州波特兰市和蒙大拿州白鱼市等风景如画的景点蓬勃发展，房价也随之上涨。在东北部和中大西洋地区，人们关注的焦点是纽约州北部，即从哈德逊河谷到宾夕法尼亚州边境的广阔土地。在2020年混乱的夏季，如果你买不起新的乡村住宅，那么退而求其次的选择就是租一套来逃离现实，而最快的方式就是通过爱彼迎。算法市场让搬家变得像点击按钮一样简单——但其速度和规模也意味着租房变得难以承受。如果你行动足够快，并且支付得起，那么地理位置就毫无障碍；如果你支付不起，你就会被困住，而且很可能面临更高的新冠感染风险。

During the COVID-19 pandemic, the same algorithmic forces that shaped tourism quickly began to shape movement within the United States as well. Within a matter of weeks, the ways in which many people evaluated real estate flipped. Dense cities had been increasingly desirable places to live, but with overloaded hospitals and the claustrophobia of quarantine, density became far less appealing. In 2020 and 2021, there was a marked population shift to suburban and rural areas, according to the US Census Bureau’s estimates. Picturesque spots like Bend, Oregon; Portland, Maine; and Whitefish, Montana, boomed, and property prices rose in turn. In the Northeast and mid-Atlantic region, the focus was on upstate New York, the broad stretch of land from the Hudson Valley to the Pennsylvania border. During the chaotic months of summer 2020, if you couldn’t buy a new countryside house, the next best thing was to rent one as an escape, and the fastest way to do so was on Airbnb. The algorithmic marketplace made moving as easy as clicking a button—but the speed and scale at which it happened meant that rentals became unaffordable, too. If you were fast enough and could pay enough, geography was frictionless; if you couldn’t, you were stuck, and likely stuck with higher risk for catching COVID.

2020年夏天，我和一群朋友在卡茨基尔预订了一间Airbnb，这样我们就可以从华盛顿特区、波士顿、纽约等不同城市赶来。我注意到，那些最符合“空域”美学风格的房子——干净整洁的白色墙壁，配备极简主义风格的家具和黄铜装置——往往预订得最快，价格也最高。我们最终在纽约州亨特找到了一栋大房子，里面摆放着直销品牌Joybird的切斯特菲尔德沙发，以及板状木质餐桌旁的黑漆摇椅。我们很幸运能订到这间房子——一位女士刚刚取消了她为期一整个夏天的租房，而我在12小时后就预订了。事实上，这栋房子坐落在一个院落里，院子里还有两间小木屋，是两个布鲁克林家庭为了逃离这座城市而连续几个月预订的。这些Airbnb房源是因为新冠疫情才存在的。“疫情期间我们忙着快速翻新，因为很多人想要长期租房，”房主迪尔德丽·巴顿后来告诉我。巴顿甚至在上传照片到房源之前就已经在平台上把这些小木屋租出去了。

In the summer of 2020, I booked an Airbnb in the Catskills with a group of friends so we could all trek in from different cities: D.C., Boston, New York. I noticed that the houses with the most AirSpace aesthetics—cleanly renovated with white walls, equipped with minimalist furniture and brass fixtures—tended to book the fastest and cost the most. My group eventually found a large house in Hunter, New York, that sported a Chesterfield sofa from the direct-to-consumer brand Joybird and black-painted Shaker chairs along the slablike wood dining table. We were lucky to get it—a woman had just canceled her summer-long rental, and I booked it twelve hours later. In fact, the house was set in a compound, with two smaller cabins on the grounds, which two Brooklyn families had booked for months on end to get out of the city. The Airbnbs existed because of COVID-19. “We went into mad pandemic time just trying to renovate them really quickly, because so many people wanted long-term rentals,” the property’s owner, Deirdre Patton, later told me. Patton had rented out the cabins on the platform before she even uploaded photos to the listings.

突如其来的移民潮导致纽约州北部地区出现了一些奇怪的现象。房屋装修仓促进行，几乎一件家具都没有就被抛到市场上，管道破损、地板破洞等问题也显而易见。正如一家租赁公司的创始人告诉我的那样，不习惯使用全套厨房的城市租房者经常会发生一些意外，比如把馅饼放在烤箱里烤糊了，从而触发火警警报。随着纽约市人口涌向北部，占据了本已不多的住房存量，一种涡轮式的中产阶级化正在发生。据一位房地产经纪人称，房屋售价比要价高出20%以上。这不仅仅是纽约州北部的情况；那些突然能够（并且被迫）在家工作的人可以在任何地方工作。选择包括墨西哥城、里斯本和巴厘岛等热门的远程办公目的地，这些地方迎来了大批新的长期访客。这些人是“数字游民”——他们在旅行的同时从事收入丰厚的远程工作或自由职业，在风景如画、相对便宜的地方用笔记本电脑工作。

The suddenness of migration led to some strange phenomena across upstate New York. Renovations were rushed, and houses were thrown onto the marketplace with barely a piece of furniture in them, with conspicuous issues like broken plumbing or holes in the floors. City renters unused to full kitchens had accidents like leaving a pie to burn in an oven and setting off the fire alarm, as the founder of one rental company told me. A kind of turbo-gentrification was happening as New York City’s population rushed northward, occupying the already low housing stock. Homes were selling for more than 20 percent over asking, according to one Realtor. It wasn’t just upstate New York; those suddenly able (and forced) to work at home could work from anywhere. The options included popular remote-work destinations like Mexico City, Lisbon, and Bali, which saw influxes of new long-term visitors. These were “digital nomads”—people who hold down lucrative remote jobs or freelance gigs while traveling, working from their laptops in picturesque, relatively affordable locations.

当然，即使没有 Airbnb，这种疫情期间的迁徙也依然会发生。但 Airbnb 平台加速并加剧了这种迁徙，因为它让租客能够轻松找到他们心仪的住所，并激励房主从一开始就将房屋投放到市场上。某种程度上，Airbnb 创造了这种即时迁徙的预期。该公司最突出的口号之一是“Belong Anywhere”，这项活动始于 2014 年。“体验一个地方，就像你在那里生活一样”是另一个口号。其潜在信息是，你不必再成为一名游客；你可以住在当地人的家里，体验他们真正的生活。

This pandemic migration would have still happened without Airbnb, of course, but the platform accelerated and exacerbated it by making it so easy for a renter to find exactly the space they were looking for, and by motivating owners to put their homes on the marketplace to begin with. In a way, Airbnb had created the expectation of such immediate movement. One of the company’s prominent slogans was “Belong Anywhere,” with a campaign that began in 2014. “Experience a place like you live there,” was another slogan. The underlying message was that you don’t have to be a tourist anymore; you can playact the life of an actual local by staying in their home.

Airbnb 由 Brian Chesky 和 ​​Joe Gebbia 于 2008 年创立，两人是罗德岛设计学院的室友兼同学，都学过工业设计，还有一位是开发人员内森·布莱卡斯亚克。大学毕业后，他们搬到了旧金山，萌生了在公寓客厅里铺个充气床垫，出租做成低保真民宿的想法。在旧金山的一次设计会议上，他们找到了第一位客户，当时酒店人满为患，游客几乎没有经济实惠的住宿选择。这种场景体现了该平台对亲密感和归属感的承诺：它本应是关于闲置空间的，是一种近乎慈善的分享行为，却又披着净社会正能量的外衣。陌生人之间可能会互相认识，而房东（或偷偷租房的人）可以通过将并非总是必要的房间变现来维持自己的经济。这种交集催生了一家如今市值超过800亿美元、拥有超过1.5亿用户的公司。Airbnb或许是规模庞大的数字社交网络，其对实体经济的影响最为直接。

Airbnb was launched in 2008 by Brian Chesky and Joe Gebbia, roommates and former classmates at the Rhode Island School of Design who had both studied industrial design, along with Nathan Blecharczyk, a developer. After college, they moved to San Francisco and had the idea of laying out an air mattress in the living room of their apartment to rent it out as a very lo-fi bed-and-breakfast. They landed their first customer during a design conference in the city, when hotels were crowded and visitors had few affordable options for overnight stays. This scenario underlies the platform’s promise of intimacy and belonging: It was supposed to be about spare space, an act of almost charitable sharing that had the veneer of net social positivity. Strangers might get to know each other, and owners (or surreptitious renters) could financially support themselves by monetizing rooms that weren’t always necessary. That intersection resulted in a company now worth over $80 billion, with over 150 million users. Airbnb might be the massive digital social network with the most direct physical consequences.

实际上，该公司构建了一个利润丰厚的市场，许多房源仅用于短期租赁——无人居住，出租公寓，甚至一些整条街道都营造出酒店的氛围，就像马克·奥杰（Marc Augé）那座失控的非场所建筑一样。在早期，Airbnb 派出专业摄影师记录房源，确保所有房源都具有相同的冷峻、广角和鲜明的审美，以便像 Instagram 一样吸引人。辅助的家庭手工业应运而生，比如 Airbnb 经理，他们在客户离开之前整理房源，以及设计顾问，他们可以就符合平台用户群体的装修提供建议。其中一位顾问娜塔莎·福伦斯（Natascha Folens）在 2016 年告诉我，业主应该接受“工业风格和中世纪风格”。“只要它看起来不杂乱和陈旧，”她补充道。

In practice, the company built such a lucrative market that many of the listings were short-term rental only—no one lived in them, lending apartments and even some entire streets the atmosphere of a hotel, like Marc Augé’s non-place run amok. In its early years, Airbnb sent professional photographers to document properties, ensuring that the listings all shared the same cold, wide-angled, aggressively bright aesthetic, the better to be appealing as Instagram-like digital images. Ancillary cottage industries popped up, like Airbnb managers, who tidied the properties between customers, and design consultants, who could advise on decoration that would appeal to the platform’s demographic. Natascha Folens, one such consultant, told me in 2016 that owners should embrace “the industrial look and the mid-century.” “As long as it doesn’t look cluttered and old,” she added.

当我回顾十年前在 Airbnb 上的个人资料时，我看到了同样空旷的白色墙壁空间：里斯本山顶的公寓，我和杰西必须支付滞纳金才能入住；塞维利亚庭院，带有引人注目的木制百叶窗；纽约州北部由模特转行做餐馆的玻璃盒子；东京的无菌阁楼工作室，它有自己的即使在密集的城市中心，我也能找到一栋小楼；在蒙马特附近，我还能找到一栋整洁的巴黎公寓，里面收藏着琳琅满目的艺术书籍。我去过所有这些地方，但从某种意义上说，它们只是一个地方，只存在于我的账户中，不属于任何地方。当然，这些真实的城市依然存在。普通人在其中生活，他们的密度和个性是任何一次旅行都无法涵盖的。但我与这些地方的接触有限，我被引导去体验那些强化我自身观点的体验，那些由我所有在线个人数据构建起来的品味档案。我感觉自己栖息在每个目的地的Instagram版本中——一个肤浅的、表面上互动的图像，最终并不能反映现实。事实上，这些图像遮蔽了现实，使现实更难被找到。这种感觉——真相和真实性难以捉摸，我们被一系列数字媒介所阻碍，不愿去寻找它们——似乎正是《过滤世界》的核心所在。

When I scroll backward through the history of my decade-old profile on Airbnb, I see grids of similarly spare, white-walled spaces: the hilltop Lisbon apartment where Jess and I had to pay a late fee to get in; the Seville courtyard with dramatic wooden shutters; the woodland glass box in upstate New York owned by a model-turned-restaurateur; the sterile lofted studio in Tokyo contained in its own tiny building even in the midst of the dense city; the immaculate Parisian apartment near Montmartre with its perfect collection of art books. I had been to all these places, but in a sense, they were one place, contained within my account, not belonging to anywhere at all. These real cities still persisted, of course. Normal people carried out their lives in them with a density and individuality that no one trip could ever encompass. But my encounter with the places had been limited, and I had been guided toward experiences that reinforced my own point of view, the taste profile built up by all my personal data online. I felt that I had inhabited an Instagram version of each destination—a shallow, superficially interactive image that ultimately did not reflect reality. In fact, the image occluded reality, made it harder to find. That feeling—that truth and authenticity are evasive and we are discouraged from seeking them by a series of digital mediations—seems to be at the heart of Filterworld.

近年来，在Airbnb占据主导地位的城市，墙壁上出现了反对Airbnb的涂鸦。雅典的涂鸦是“停止Airbnb和旅游业”和“我们的城市变成了商品”。威尼斯的涂鸦是“去他妈的Airbnb”。葡萄牙科英布拉的涂鸦是“一个Airbnb游客会把两三个学生赶出我们的城市”。（这些都是对算法以及美国硅谷公司整体主导地位的抗议。）巴塞罗那和柏林因对Airbnb的严格监管而闻名，限制了谁可以在该平台上出租房屋以及出租期限。人们对Airbnb这家公司产生了强烈反对，抗议其高昂的收费和对不可靠房东的宽松审批。也许酒店，这种最初千篇一律的体验，正在重新流行起来？但我从未放弃过Airbnb。我被那种抽象的承诺深深吸引，想要在几天内体验一种与自己截然不同的真实生活，想象自己是里斯本的雕塑家，或者东京的音乐家。对我来说，如果我住在真正的公寓里，旅游的幻觉就会更加彻底——尽管这当然仍然是一种幻觉。

In recent years, anti-Airbnb graffiti has popped up on the walls of cities where it is a dominant force. “Stop Airbnb and tourism” and “Our cities turn into commodities” in Athens. “Fuck Airbnb” in Venice. “One Airbnb tourist kicks out 2 to 3 students out [sic] of our city,” in Coimbra, Portugal. (These were protests against algorithms as well as the overall dominance of American Silicon Valley companies.) Barcelona and Berlin became known for strict regulation of Airbnb, limiting who could rent out their spaces on the platform and for how long. There was a backlash against Airbnb as a company, protesting its aggressive fees and lax approval of unreliable hosts. Maybe hotels, the original generic experience, were coming back into fashion? But I’ve never given up Airbnbs. I’m too attracted by that abstract promise, to live an authentic life that’s different from your own for a few days, imagining being a sculptor in Lisbon, say, or a Tokyo musician. For me, the illusion of tourism is more total if I stay in a real apartment—though of course it remains an illusion.

Airbnb 作为一家公司，深知其扁平化效应。2021 年末，我与 Airbnb 现任首席执行官 Chesky 进行了交谈。那是一次 Zoom 电话会议；Chesky 当时在他旧金山的家中，尽管背景被滤镜模糊了。他语速很快，语气坚定，仿佛一个早已心知肚明的人。在所有事情上。“我正在努力吸取过度旅游的教训。我本身并不认为世界上存在过多的旅游业；我认为过度旅游主要是太多人在同一时间前往同一个地方，”他说。“如果你能设计一幅完美的画面，你会在多个日期将游客平均分配到多个地方，这样就不会让任何一个地方的游客超负荷。”切斯基出生于纽约州北部，尽管那里不像城市居民蜂拥而至的典型哈德逊地区那样充满浪漫气息：“虽然没有多少人去我的小镇，但那里其实很不错。”

Airbnb as a company is aware of its flattening effects. In late 2021, I spoke with Chesky, now the CEO of Airbnb. It was a Zoom call; Chesky was at his home in San Francisco, though the background had been blurred out into nothingness by a filter. He spoke quickly and implacably, the tone of a man who already knows his thoughts on everything. “I’m trying to learn the lessons of overtourism. I don’t per se think there’s too much tourism in the world; I think overtourism is mainly too many people going to the same place at the same time,” he said. “If you could design the perfect picture, you would fairly equally distribute people across many places across many dates, and you wouldn’t overload any one place.” Chesky was born in upstate New York, though a less romanticized area than the archetypal Hudson area that city dwellers flocked to: “Not a lot of people are going to my town, but it’s actually pretty nice.”

Chesky 表示，Airbnb 专注于解决“旅行再分配”问题。由于科技平台很少直面自身造成的巨大社会变革，而是倾向于淡化自身影响力，因此，他承认 Airbnb 不仅影响了人们的旅行方式，还影响了人们的目的地，积极地影响着目的地，这让我感到惊讶。“我们将需求导向有供应的地方，导向需要供应的城市，”他说道。与 Filterworld 中的其他内容一样，地理信息也面临着发现难的问题。当选择如此之多时，很难找到一个特定的旅行地点；推荐必须提供有趣的发现，同时也要推荐熟悉且可识别的地点。用户倾向于遵循预先设定的路线，正如 Chesky 所解释的那样：“Airbnb 上有十万个城镇。人们不会记住十万个目的地；他们只会记住大约十个，而这十个目的地是他们在 Netflix 上看到的。每个人都想去巴黎，因为《艾米丽在巴黎》。 ”

Chesky said that Airbnb was focused on solving this problem of “travel redistribution.” Since tech platforms rarely address the enormity of the social changes they have caused, preferring to deemphasize their own power, it surprised me that he admitted Airbnb had an effect not just on how people travel, but on where they go, actively influencing the destination. “We point demand to where we have supply, to cities that want it,” he said. Like other kinds of content in Filterworld, geography suffers from a problem of discovery. It’s hard to find a particular place to travel when there are so many options; recommendations must serve up interesting finds while also suggesting what is familiar and recognizable. Users tend to stick to predetermined paths, as Chesky explained: “We have one hundred thousand cities and towns on Airbnb. People don’t keep one hundred thousand destinations in their head; they keep about ten, and the ten are the ones that they see Netflix shows about. Everyone wants to go to Paris because of Emily in Paris.”

我问切斯基，这么多人去同一个地方，通过他的数字平台，他们的选择和品味受到当地风土人情的影响，他认为这会带来什么后果。“你触及了我们这个时代的核心问题之一：我们与地方的关系是什么？在一个人们可以自由跨越国界的世界里，民族主义是更多还是更少？”他问道。切斯基最终认为这种转变是积极的：“我内心的乐观主义者希望这一切意味着世界感觉更小一些。如果你能让物理世界变得更小，那几乎完全是一件好事。我认为这就是最终目标。”然而，这种“小”的愿景却在Airbnb的支持和控制之下，遵循着公司设定的模板和盈利条件。规模小意味着同质化，并趋向统一，这很可能是西方默认的做法，也是科技行业意识形态的体现。切斯基暗示的权衡是，运动越多，身份认同就越少。身份认同也关乎内容。

I asked Chesky what he thought the consequences were of so many people traveling to the same places, routing through his digital platform, their choices and tastes influenced by what they found there. “You hit on one of the central questions of our time: What is our relationship to place? Is there more or less nationalism in a world where people can freely move across countries?” he asked. Chesky ultimately sees the shift as positive: “The optimist in me hopes that the implication of all this is that the world feels a little bit smaller. If you can make the physical world smaller, that’s almost entirely a good thing. I think that’s the endgame.” Yet that vision of smallness is under the auspices and control of Airbnb, following the template the company has set and the conditions that it profits under. Smallness implies homogeneity and a move toward uniformity, most likely under a Western default, under the ideology of the tech industry. The trade-off that Chesky implied was that the more movement there was, the less identity there would be. Identity, too, is a matter of content.

追逐点赞

CHASING LIKES

对我来说，互联网一直与劳动密不可分。网络生活是工作与娱乐的混合体，而且常常两者兼具——考虑到互联网的发展，这在意料之中，因为它本身就将互联网上涌现的文化商品化了。甚至在我上大学之前，社交媒体就已经是成年人职业素养的标志：2006年夏天，只有拥有大学邮箱地址才能登录Facebook，而录取的学生直到大一入学前的那个夏天才能获得这个邮箱地址。所以，当我终于拥有一个Facebook账号时，它主要是一个社交平台，用来认识塔夫茨大学即将和我同班的其他同学。我们很快为2010届毕业生创建了一个Facebook小组，在当时看来，这似乎还很遥远。我们列出了各自的家乡，讨论了各自的专业方向，并计划了现实生活中的聚会。由于我们共同的联系，社交网络提前将我们聚集在一个数字空间里——这种社会组织形式在当时感觉很罕见，它将互联网与线下生活融为一体。那个Facebook群组促成了一件影响深远的人生大事：在那里，我和伴侣杰西第一次互动，我们开始互相留言，分享彼此共同的音乐品味。

For me, the Internet has always been inextricable from labor. Life online is an admixture of work and entertainment, very often both at once—which is predictable given that the Internet evolved to commodify the culture that emerged on it. Even before I went to college, social media was a marker of adult professionalism: in the summer of 2006, you could get on Facebook only once you had a college email address from whichever school you were going to attend, something accepted students didn’t receive until the summer before their freshman year. So when I finally got a Facebook account, it was primarily a space for networking, getting to know the other students who were going to be part of my class at Tufts University. We quickly formed a Facebook group for the graduating class of 2010, which seemed at that point far in the future. We listed our hometowns, debated what we wanted to major in, and planned IRL meetups. Because of our shared affiliation, the social network had brought us together in advance, in a digital space—a form of social organization that felt rare at the time, merging the Internet and life offline. One very consequential life event emerged from that Facebook group: it was where I first interacted with my partner, Jess, and we began messaging about our shared taste in music.

大学三年级的时候，互联网也为我提供了一些从事我理想职业的领域——写作——的有偿工作机会。我曾为一些艺术评论撰稿。我曾在塔夫茨日报举办过展览，并于2008年加入Twitter。当时，Twitter更像是一家人烟稀少的咖啡馆，而不是一座喧闹的体育场。当时和现在一样，Twitter最沉迷的用户群是记者，他们忍不住要发推文，就像狼忍不住要嚎叫一样。Twitter让我第一次接触到媒体行业和新兴的在线出版世界。（埃隆·马斯克可能在2023年将Twitter的正式名称改为X，但最初的名称在我脑海中已经根深蒂固。）

During my junior year of college, the Internet also provided some of my first opportunities for paid work in the field I wanted to make a career in: writing. I had contributed reviews of art exhibitions to the Tufts Daily newspaper, and I joined Twitter in 2008, when the platform was more like a sparsely populated café than a roaring stadium. Then as now, Twitter’s most addicted user base is journalists, who can’t help but tweet the way wolves can’t help but howl. Twitter gave me my first exposure to the media industry and the nascent world of online publishing. (Elon Musk may have changed Twitter’s official name to X in 2023, but the original is far too set in my mind.)

与此同时，当时仍以纸质杂志为主的《大西洋月刊》最近在其网站上推出了主题博客，涵盖文化和政治等广泛领域。尽管包括塔-纳西斯·科茨在内的个人作家都维护着自己的主题博客，收集他们对特定主题的想法，但这些垂直博客也刊登了许多不同的作家的作品，包括自由撰稿人。这些自由撰稿人往往是新兴的年轻作家，因为他们必须愿意为纯网络媒体撰稿，并且渴望接受非常低的稿费。一位曾在塔夫茨大学暑期实习的朋友把我介绍给了一位编辑，她接受了我的一些投稿。这些投稿都是关于文化新闻和潮流的小讨论，也就是现在所谓的“热门话题”，因为它们更多地基于昙花一现的观点，而非报道或研究。我为《大西洋月刊》撰写了一篇关于为什么一位创作型歌手比另一位更优秀的文章；后来，我为《卫报》评论了手提袋的性别象征意义，为《Vice》收集了病毒式传播的艺术作品，并为《新共和》颂扬了网络友谊的益处。这些作品的出现和消亡都很快，有些在 Facebook 或 Twitter 上引发了讨论，而有些则湮没无闻。

Meanwhile, The Atlantic, which was still primarily a print magazine at the time, had recently launched themed blogs on its website, covering broad areas like culture and politics. Though individual writers including Ta-Nehisi Coates maintained themed blogs of their own to gather their thoughts on specific subjects, these verticals published many different writers, including freelance contributors. The freelancers tended to be emerging young writers, since they had to be willing to write for online-only media and hungry enough to accept very low fees. A friend at Tufts who had interned for The Atlantic one summer introduced me to an editor, and she accepted a few of my pitches. They were little arguments about cultural news and trends, what would now be called “hot takes,” since they were based more on ephemeral opinion than on reporting or research. For The Atlantic, I wrote about why one singer-songwriter was better than another; later I weighed in on the gender symbolism of tote bags for the The Guardian, collected artworks that had gone viral for Vice, and extolled the benefits of online friendship for The New Republic. The pieces lived and died quickly, some sparking discussion in Facebook or Twitter feeds and others sinking without a trace.

回顾这些早期的档案，我并不感到自豪，但这正是当时数字生态系统所要求的。读者喜欢热门话题，因为它们发人深省，能够提供一些观点，读者可以据此形成自己的观点；编辑和作家也欣赏它们，因为它们写得快，委托成本低，而且易于在线发布。如果你写得足够快，一个下午就能从零开始写到出版。我为此获得的报酬虽然微不足道，但仍然令人向往，只有几百美元（《大西洋月刊》的起价是五十美元）。就像二十世纪的作家看到自己的第一篇作品付梓时那样，激动不已。我的署名以像素级显示于《大西洋月刊》的数字横幅下，并采用其标志性字体。这赋予了文本一种权威性，而当我独自一人在波士顿学生公寓的客厅里用Word文档起草时，我从未想过它会拥有这种权威性。（数字空间就像普鲁斯特的玛德琳蛋糕一样唤起回忆：在我写下这些文字的时候，我仿佛能嗅到静止炎热的夏日空气、明媚的阳光，以及我写作和编辑我的第一篇小说时所坐的宜家沙发的硬垫。）当然，我也在推特上分享了我的文章，在那里我关注了一群主要从事艺术界的新兴群体。这是我第一次接触当代持续不断的网络自我推广。你不仅要发表一些东西，还必须同时推销作品和你自己。

Looking back on this early archive doesn’t exactly fill me with pride, but it was what the digital ecosystem of the time demanded. Readers like hot takes because they’re voicey and provide an opinion against which one can shape one’s own; editors and writers appreciate them, too, because they’re fast to write, cheap to commission, and easy to publish online. If you worked fast enough, you could go from zero to published in an afternoon. For these I was paid a pitiable but still-desired fee, in the low hundreds of dollars (The Atlantic started out at fifty dollars). The way twentieth-century writers must have felt about seeing their very first piece run in print, it was a thrill to have my byline appear in pixels under The Atlantic’s digital banner and in its signature typeface. It gave the text an authority that I didn’t know it could have when I was drafting it in a Word document alone in the living room of my student apartment in Boston. (Digital spaces evoke memories just like Proust’s madeleine: as I write this, I’m getting a whiff of the still, hot summer air, the bright sun, and the stiff cushions of the IKEA couch where I wrote and edited my first stories.) Of course, I also shared my articles on Twitter, where I followed a nascent crowd who were mostly engaged with the art world. It was my first brush with the constant contemporary labor of online self-promotion. Not only did you have to publish something; you had to market the piece and yourself at the same time.

人气竞赛是文化中一个永恒的现象。（在史前时期，肯定有一些洞穴壁画比其他壁画更受追捧，甚至引发了尼安德特人排长队逐一参观的现象。）中国古代绘画的收藏家习惯于在心仪的画作上盖章，盖上自己的名字——类似于在推特上点赞或点赞。几个世纪以来，杰作会获得大量的印章，有些甚至会盖在山水画上。在西方艺术中，一幅画作周围镶有徽章的精美画框或许能彰显其重要性，书籍或专辑上“白金销量”的畅销书标签也同样如此——所有这些都象征着极高的文化价值，或者至少是极高的经济价值。这些信号始终引导着消费者可能购买并欣赏的商品。例如，畅销书标签可能会鼓励购物者再看一眼书籍，或者翻到第一页。艺术家一直以来都必须扮演某种营销者的角色，不断塑造公众形象，作为其作品的名片，就像杰克逊·波洛克和安迪·沃霍尔那样。如今的不同之处在于，在“过滤世界”（Filterworld）中，点赞数量、预先存在的关注度等指标往往比文化作品本身更有说服力。这些指标不仅是成功的衡量标准，更是创造成功的要素，因为它们决定了作品首先会被推荐给观众，最终观众会看到什么。这就像12世纪的中国艺术家必须获得足够多的印章才能展出自己的作品一样。

Popularity contests are a perennial feature of culture. (In prehistory, surely there were cave paintings that were hyped up more than others and inspired Neanderthal queues, one viewer at a time.) Collectors of ancient Chinese painting habitually added their names to favored paintings by stamping them with seals—akin to an approving tweet or a thumbs-up like. Masterpieces acquired bunches of stamps over centuries, with some even layered on top of the painted landscape itself. In Western art, an elaborate frame with an insignia around a painting might speak to its importance, as would a bestseller-list label on a book or an album “going platinum”—all symbols of a high cultural value, or at least a high economic appraisal. These signals have always guided what consumers are likely to consume and thus appreciate. The bestseller label, for example, might encourage a shopper to give a book a second glance or flip to the first page. Artists have always had to be marketers of a sort, evolving public personas that served as calling cards for their work, as Jackson Pollock and Andy Warhol did. The difference today is that in Filterworld, the metrics—the number of likes, the preexisting attention—tend to speak louder than the piece of culture itself. Not only do they act as a measure of success, but they create success, because they dictate what is recommended to and seen by audiences in the first place. It’s as if a Chinese artist in the twelfth century had to acquire a large enough number of stamps before he could show his work.

我很快就学会了在互联网上判断我的成功单纯从数量上衡量用户，这种现象在主流社交媒体的早期阶段激增。Facebook 帖子或推文的点赞数量代表了它触达了多少人，有多少人被你的自我宣传公告或观点文章所激励而点击按钮。但这真的能代表帖子的质量吗？我们应该用这个武断的新指标来评价自己吗？这些问题迫在眉睫，但考虑到新社交网络的实用性及其带来的即时覆盖能力，人们很容易忽略这些困境，继续发帖。

I quickly learned to judge my success on the Internet almost solely in terms of numbers, which proliferated in the early stages of mainstream social media. The number of thumbs-ups on a Facebook post or Tweet was a representation of how many people it reached, how many people were inspired to click a button by your little self-promotional bulletin or opinionated missive. But was that a representation of a post’s quality? Should we evaluate ourselves by this arbitrary new metric? The questions loomed, but given the utility of new social networks and the instant reach they afforded, it was easy to ignore such quandaries and just keep posting.

突然之间，用户可以比较每一条内容与其他内容的优劣。数字无处不在，它会在你阅读或观看的所有内容旁边弹出。即使点赞、分享、转发或收藏的数量是平台本身强加的人工价值衡量标准，它们仍然成为判断内容在海量信息流中脱颖而出的最快、最简单的方法。越多越好：如果一个YouTube视频的观看次数很高，或许它很有见地，或者至少很有趣。病毒式传播等同于质量——其他人喜欢，所以你也会喜欢。对于创作者来说，无论是作家还是有抱负的Instagram网红，指标都是目标、动力和内在的指南针，用来评估哪些有效、哪些无效。尤其是在新闻行业，数字越高，某件事似乎就越重要。这是一个不言而喻的道理：点赞越多意味着更多人看到了它。在看似精英至上的互联网上，每一个点赞都是一次投票，任何人都可以上传内容，但肯定不是每个人都能激发观众的关注。

Suddenly, users could tell how every little piece of content compared to every other one. The numbers could not be escaped, popping up next to everything you read or watched. Even if the count of likes, shares, retweets, or faves was an artificial measure of value imposed by the platform itself, they still became the fastest and easiest way to judge what popped up in the infinite feeds of content. More was always better: if a YouTube video had a lot of views, perhaps it might be insightful or at least funny. Virality equaled quality—other people liked this, so you will, too. For creators, whether writers or aspiring Instagram influencers, metrics were the goal, the incentive, and the internal compass evaluating what worked and what didn’t. Particularly in journalism, the higher the numbers, the more important something seemed to be. It was a truism: more likes meant more people had seen something. Each like was a vote in the seemingly meritocratic Internet, where anyone could upload content but certainly not everyone could inspire audiences to pay attention.

病毒式传播法则——点赞是算法判断特定内容值得更多推广的指标——在数字媒体早期尤为重要。当时，出版物网站开始在文章标题附近添加快捷按钮，方便在Facebook、Twitter以及越来越多的其他平台上分享文章。后来，这些按钮也更新了，带有专属的小数字，显示文章在读者看到内容之前就获得了多少点赞或转发。点赞成了互联网的货币，记者们被迫追逐点赞。（当然，我们这样做并不需要太多的鼓励。）

The law of virality—likes as an indication to the algorithm that specific content deserved more promotion—was particularly relevant for the early days of digital media. Publications’ websites began to add shortcut buttons near the headlines of stories to share an article on Facebook, Twitter, and a proliferating list of other platforms. Then the buttons were updated to have their own tiny numbers, showing how many likes or retweets an article had gotten even before the reader had encountered its contents. Likes became the currency of the Internet, and journalists were forced to chase them. (Not that we needed much encouragement to do so.)

作为一名作家，我密切关注着我文章的点击量。我当我的故事比别人获得更多的点赞时，我会感到自豪。当我在推特上发布链接或在脸书上分享我的故事时，我会注意到哪些帖子得到的回复最多。最成功的故事往往是那些影响读者个人生活，或具体化他们所面临的抱怨或问题的故事。当我发表权威或戏剧性的言论时，得到的回应也是巨大的。我会通过我们永远在线的 Gchat 窗口招募朋友和同事，让他们在文章发布时立即点赞，以此来为我的成功铺平道路。多年后，当这些按钮不再受欢迎，可见的数字从文章页面上消失时，我松了一口气。但在过去的十年里，每当我发表一篇文章时，我都会担心到底有多少人点击了点赞和转发。我可以尝试钻研这个系统，并了解哪些方法可能会在社交媒体上产生额外的参与度，但最终结果是我无法控制的。

As a writer, I closely monitored the clicks on my articles. I was proud when my stories netted more likes than others’. When I tweeted links or shared my stories on Facebook, I noticed which posts got the most responses. The most successful tended to be the stories that impacted readers’ personal lives or crystallized a complaint or problem they were having. When I made authoritative or dramatic statements, the reaction was dramatic in turn. I would recruit friends and colleagues via our always-on Gchat windows to prime my success by liking something right when it went up. It was a relief, years later, when these buttons fell out of favor and the visible numbers disappeared from article pages. But every time I have published an article over the past decade, I have fretted over just how many people were hitting Like and Retweet. I could try to game the system and be aware of which approaches might create extra engagement on social media, but ultimately the result was out of my control.

点赞的暴政在某种程度上是我们所处的网络算法生态系统的产物。每一次点击都是一个需要被吸收的数据信号，一个表明你正在关注某条内容的小信号，随后可以用来训练机器学习系统，为你提供更多类似的内容。计算机不会追踪我们的眼球运动——至少目前还没有——因此像点赞或表情符号这样的按钮，就成了我们对内容反应的替代物。这些是自我表达的便捷捷径，但不可避免地也为定向广告提供了更多素材。我们如此公开地表达自己的偏好，其实是在参与自我监控。

The tyranny of likes is in part a function of the algorithmic ecosystem we exist in online. Each button press is another data signal to be absorbed, a small sign that you’re paying attention to a piece of content that can then be used to train machine-learning systems to serve you more of that thing. Computers don’t track our eye movements—at least not yet—and so buttons like the thumbs-up or emoji reactions serve as a proxy for our reaction to content. These are convenient shortcuts for self-expression but also, inevitably, provide more grist for the mill of targeted advertising. We participate in our own surveillance by signaling our preferences so publicly.

衡量点赞的不仅仅是文章。Instagram 上的自拍、Facebook 上关于假期或婚礼的状态更新，以及 Medium 上宣布新商业计划的帖子也是如此。Medium 是一个极简主义的博客平台，由 Twitter 的一位联合创始人于 2012 年创立，曾一度采用“点赞”作为其衡量标准。用户可以按下按钮为文章点赞，并根据自己的喜好随意点赞——最多 50 次。有一段时间，点赞的数量也决定了 Medium 向使用其平台的作者支付的费用，这实际上是点赞与价值的等式。随着时间的推移，点赞出现了一种膨胀的现象。每一个点赞的价值都越来越低。到了 2020 年，对某篇文章点赞与其说是喜欢它，不如说是感知到了它——它已经突破了整体互联网的喧嚣。记者们常常在社交媒体个人简介中加上“点赞不等于认可”的附加条款，以防止被误解。虽然你并不总是因为喜欢某个帖子才点赞，但点赞是你唯一可用的功能。

It wasn’t just articles that were measured by likes. So were Instagram selfies, Facebook status updates about vacations or weddings, and Medium blog posts announcing new business ventures. Medium, a minimalist blogging platform launched by a cofounder of Twitter in 2012, at one point adopted “claps” as its governing metric. Users could press a button to clap for a piece and click it as many times as they were inclined to—up to fifty claps. For a time, the number of claps also dictated the fees that Medium paid to writers who used its platform, a literal equation of likes to value. Over time, a kind of inflation of likes occurred. Each one had less and less value. By 2020, hitting Like on something wasn’t so much a sign that you enjoyed it as just that you had perceived it—it had broken through the overall noise of the Internet. It was common for journalists to put a proviso in their social media bios reading “Likes don’t equal endorsements” to guard against misinterpretation. Though you don’t always press the Like button because you actually enjoy a post, liking is the only function available to you.

每个社交媒体用户都对什么会获得点赞有着与生俱来的感觉。如同古典美或几何比例一样，这个公式并不精确，却总能被识别。挑衅会激发点赞，因为点赞是一种效忠和赞同的姿态，象征着用户是站在喷子一边还是被喷子一边。愤怒会获得点赞，因为点赞传达了同情的愤怒：你也很生气，因为怎么可能有人不生气呢？因此，一种常见的帖子格式是声称没有足够多的人谈论某个话题：一场战争、一场灾难、一项糟糕的政策、一个坏人。单是这种无知就足以令人愤慨！性话题也会获得点赞，原因无需解释。政客泰德·克鲁兹的推特账户就曾点赞过一段色情视频，那段性行为被归咎于一名工作人员。（2015年，一篇文章在标题中发出警告：“给男士们的公益广告：每个人都能看到你在Instagram上点赞少女的比基尼照片。”）

Every social media user has an innate sense of what will get liked. As with classical beauty or geometric proportion, the formula is inexact and yet always identifiable. Provocation inspires likes, since the like is a gesture of allegiance and agreement, a symbol of whether the user is on the side of the troll or the trolled. Outrage gets likes because the likes signal sympathetic outrage: You are mad, too, because how could anyone not be? Hence a common format of post claiming that not enough people are talking about a given subject: a war, a disaster, a bad policy, a bad person. The ignorance alone is an outrage! Sex gets likes, for reasons that don’t require explaining. The politician Ted Cruz’s Twitter account once liked a pornographic video, a lusty act that was blamed on a staffer. (In 2015, one article offered a warning in the headline: “A PSA for Dudes: Everyone Can See You Liking Teen Girls’ Bikini Pics on Instagram.”)

此外，还有一些更容易被接受的情绪可以用来引人点赞。幽默很有效，既因为你喜欢这个笑话，也因为你想把它传递给别人，所以分享会引发更多分享。关联性之所以有效，是因为它能让社交媒体上尽可能广泛的受众从某件事中认出自己。相关的内容包括饮食习惯、对自己懒惰的抱怨，以及常见的童年经历。（“你有没有……？”）我曾经发过一条推文：“抱歉我回复晚了，我太害怕打开它了，这毫无道理。”这番话是我对邮箱焦虑的短暂反思。这句话既不特别有趣，也不特别有说服力，但有近15万人点赞——在某种程度上，他们点赞就是在表达自己的想法。一位作家朋友习惯性地把她最平庸的观察留到周末发推文，因为那时观众们正在寻找特别能引起共鸣的内容。这通常很有效，经常能收获成千上万的点赞。最后，从怀旧或熟悉的角度来说，可识别性是提升点赞和参与度的绝佳方式。你喜欢是因为你知道事实是，你更有可能喜欢之前看过很多次的电视节目片段，而不是你没看过的电视节目片段，因为观看电视节目片段需要你坐下来仔细观看并进行评价。

Then there are more acceptable emotions used to elicit likes. Humor plays well, both because you like the joke and because you want to pass it along to others, so sharing begets more sharing. Relatability works because it allows the widest possible audience across social media to recognize themselves in something. Relatable content includes eating habits, complaints about your own laziness, and common childhood experiences. (“Did you ever…?”) I once tweeted, “sorry I responded late to your email I was unjustifiably terrified of opening it,” a momentary reflection on my own inbox anxiety. The line was neither particularly funny nor eloquent, but nearly 150,000 people liked it—by liking it, they were saying it themselves, in a way. A writer friend habitually saved her most banal observations to tweet on weekends, on the basis that it was then when audiences were looking for something particularly relatable. It often worked, regularly netting tens of thousands of likes. Finally, recognizability, in the sense of nostalgia or familiarity, is perfect for driving likes and engagement. You like because you know what something is; you are more likely to like a clip of a TV show that you’ve seen many times before than a clip of one you haven’t seen, which would require sitting through the video and evaluating it.

这些激发点赞的情感是我们人类最基本的本能，这很合理，因为它们也必须在我们继续滚动屏幕之前的一瞬间发生。这里容不下模棱两可、微妙或不确定——你不能按下按钮来表明你仍在思考某事，还没有得出任何简单的结论。绝对才是王道。Twitter 长期以来一直是社交媒体上最热闹的舞台，人们的活动归结为每天的点赞竞赛。也许是因为空间太小——一次只有几句话——而且互动方式也非常有限：转发、点赞、回复、忽略。在古罗马角斗士比赛中，皇帝用手势决定失败者的命运，竖起大拇指表示死刑，拇指按下表示宽恕。在Twitter上，聚集的用户构成了一座隐喻性的竞技场，我们都实时举手，用点赞来评判。（Facebook仍然使用竖起大拇指的图标，尽管它不再意味着死亡。）我们点赞某人对另一个人的猛烈抨击，因为我们认同这种批评。我们喜欢一个机智的反驳——一个“扣篮”——因为我们认同被批评者的观点。而每一个动作都通过算法信息流得到强化，并分发给其他人，让他们自己做出判断。

These like-inspiring emotions are some of our basest human instincts, which makes sense because they also must occur in the space of an instant, before we continue scrolling. There is no room here for ambiguity, subtlety, or unsurety—you can’t hit a button to show that you’re still mulling something over, that you haven’t come to any easy conclusions yet. Absolutes are the order of the day. Twitter has long been the feistiest arena on social media, where the action boils down to a daily contest for likes. Perhaps it’s because the space is so minimal—only a few sentences at a time—and the methods of interaction so limited: retweet, like, reply, ignore. In ancient Roman gladiatorial contests, the emperor dictated the loser’s fate with a hand gesture, thumbs up for a death sentence and thumb pressed down to the fist for mercy. On Twitter, the assembled users fill the metaphorical colosseum, and we’re all raising our hands in real time to pass judgment with our likes. (Facebook still uses a thumbs-up icon, though it no longer means death.) We like someone’s broadside against another person because we agree with the critique. We like a smart rejoinder—a “dunk”—because we agree with the person being critiqued. And each action is reinforced by the algorithmic feed, distributed to others to decide their own judgments.

点赞的暴政如此彻底，以至于艺术家尼克·德马科（Nick DeMarco）——他曾是2010年代初期互联网艺术圈的一员——在2016年发明了一款名为“Instagram上0个点赞”的游戏。这是对社交媒体常规规则的达达主义式颠覆：你不必发布最吸引人、最引人注目的照片，而是必须发布尽可能中性的照片。它甚至不能太丑，因为那样会引起太多反响，或者你的粉丝可能会讽刺地喜欢它。“这个游戏看似简单，实则无限复杂，”德马科在一套官方规则中写道。我当时也亲自尝试过；在Instagram规模较小、你知道肯定会有几个朋友看到你的帖子的时候，这甚至更难，因为在信息流过于算法化、被太多品牌和陌生人的账户占据之前。完全平凡的效果最好，比如一张不协调的人行道照片或一面随机墙壁的照片。有时我能得到一个点赞，但永远不会。真正的目标是没有。德马科提出了一个替代方案，他和朋友在一天内发布帖子，并尝试获得最低的点赞数，就像打高尔夫球一样：“获胜者取决于参赛者分开时点赞数最少的人。”这是一个有趣的任务，但它也强调了互联网上所有活动在多大程度上都是为了获得更多点赞的目标而进行的。

The tyranny of likes is so complete that in 2016 the artist Nick DeMarco, who was part of an Internet art scene in the early 2010s, came up with a game called “0 Likes on Instagram.” It was a Dadaist reversal of the usual rules of social media: instead of posting the most attractive, compelling photo, you had to post as neutral an image as possible. It couldn’t even be ugly, since that would receive too much of a reaction, or your followers might like it ironically. “The game is deceptively simple yet infinitely complex,” DeMarco wrote in a set of official rules. I tried it myself at the time; it was even harder when Instagram was smaller and you knew a few friends would certainly see your posts, before the feed was too algorithmic and occupied by too many accounts from brands and strangers. Total mundanity worked best, like an off-kilter photo of the sidewalk or a random wall. Sometimes I could get a single like, but never the true goal of none. DeMarco proposed an alternative of posting with a friend over the course of a day and trying to get the lowest number of likes overall, like golf strokes: “A winner is determined by whoever has the fewest likes at the point when the competitors split up.” It’s an amusing quest, but it also underlines the extent to which all activity on the Internet is driven toward the goal of more and more likes.

2008年，我因为看到一则新闻开始使用Twitter，报道一名美国​​留学生在埃及被捕，并用Twitter通知了家人。当时我即将前往中国留学，所以觉得Twitter或许会派上用场。Twitter的确有用，但原因并不那么戏剧化。它让我了解家乡朋友们的动态，以及新兴的网络艺术圈。很长一段时间里，Twitter一直是我生活中不可分割的一部分，像一个持续不断的职场闲聊平台，让我的日子充满了各种闲聊。对于一个非自由职业者来说，这个角色或许应该由办公室同事来承担。然而，随着时间的推移，情况发生了变化；Twitter似乎占据了越来越重要的精神空间，既承载着当下的文化讨论，也推动着这些讨论。2016年，超级活跃的Twitter用户唐纳德·特朗普当选总统，或许是一个转折点。算法推送不仅变得越来越普遍，社交网络也成为了“抵抗”阵营（反对特朗普的松散自由派联盟）舆论战的温床，他们在某种程度上也继承了右翼对网络喷子的偏爱。特朗普对推特的使用，直到2021年国会骚乱后被封禁，也加剧了其影响力：算法推送确保尽可能多的用户看到他的推文，即使他们讨厌推文的发起者。推特不仅仅是包含新闻；它本身就是新闻。

I got on Twitter in 2008 because I saw a news story that an American student studying abroad had been jailed in Egypt and had used the site to alert his family. I was about to study abroad in China, so I figured it might prove useful. It did, but not for so dramatic a reason. It kept me in touch with what my friends back home were doing, and the nascent online art world. For a long time, Twitter remained a segmented part of my life, a constant professional water cooler that filled my days with ambient chatter, a role that might have been filled by office colleagues for a non-freelancer. Something shifted over time, however; Twitter seemed to occupy an increasing amount of psychic space, to both play host to and drive the cultural discussions of the day. The 2016 election of Donald Trump, a hyperactive Twitter user, as president was likely the turning point. Not only were algorithmic feeds becoming more pervasive, but social networks became host to the parallel public-opinion battles of “the resistance,” the loose liberal coalition against Trump that adopted, to an extent, the right’s predilection for online trolling. Trump’s use of Twitter, until he was banned in 2021 after the Capitol riot, only intensified its presence: the algorithmic feed ensured as many users as possible saw his tweets, even if they hated their originator. Twitter didn’t just contain the news; it was the news.

这个平台的引力无情地塑造了我的写作内容和写作方式。当我用它来推广我的写作，并测试我的想法，看看读者的反应时，我学会了什么方法有效，什么方法无效。权威的言论、俏皮话和对某一问题任何一方的挑衅性论点都大受欢迎。我们这些2010年代中期的推特用户，发展出了一套关于引用和缩写的暗语，塑造了该平台的实时讨论。在我使用推特的高峰期，当一个想法浮现在我的脑海中时，它潜意识里就是为推特量身定制的，限制在假设的280个字符以内。我当时不再意识到思想和推文之间的转换；我的大脑已经完全接受了社交媒体关注的多巴胺刺激，就像巴甫洛夫的狗听到铃声就会流口水一样。

The platform’s gravity inexorably shaped what and how I wrote. As I used it to promote my writing and test out my ideas to see how audiences responded, I learned the mechanics of what worked and what fell flat. Authoritative statements hit, as did one-liner jokes and provocative arguments on any side of an issue. We Twitter users of the mid-2010s developed an argot of references and shorthands that shaped the platform’s real-time discourse. During my peak periods of usage, when an idea came into my mind, it arrived subconsciously made for Twitter, restricted to a hypothetical 280 characters. I was no longer aware of the translation between thought and tweet; my brain had been completely trained on the dopamine hits of social media attention, like Pavlov’s dog salivating at the ringing of a bell.

点赞并非唯一的奖励；它们存在于更广泛的线上注意力经济中，并渗透到整个线下经济。点赞带来关注。关注带来新的粉丝；粉丝们反过来点赞并分享我的作品。更多的粉丝让我获得了一层个人权威的假象：我在我的领域里是一个值得倾听的声音，一个值得关注的作家。而这份名声让我获得了编辑的委托、兼职和全职工作，这让我回到了那个循环的起点。获得更多点赞感觉就像我应该做的事情；感觉就像工作一样，而我的工作做得越来越好。

The likes were not the only reward; they existed in a wider online attention economy that bled into the offline economy at large. Likes lead to attention. Attention leads to new followers; followers who liked and shared my work in turn. More followers led to a veneer of personal authority: I was a voice to listen to in my field, a writer worth looking out for. And that reputation got me commissions from editors, part-time gigs, and full-time jobs, which drove me back to the beginning of that loop. Getting more likes felt like what I was supposed to be doing; it felt like work, and I was getting better at my job.

除了个人在社交媒体上的存在感之外，企业和文化产品也必须吸引点赞。时尚品牌可能会以Instagram点赞数量来衡量其成功，因为更多的点赞意味着更多的客户参与度和潜在买家。新电影或电视节目的账号同样会以点赞数量为导向，因为点赞是衡量营销成功的一个指标：如果你能为《暮光之城》续集《新月》的账号获得更多粉丝，那么就会有更多粉丝为之兴奋，这部剧的首播也会吸引更多观众。这就是赢得喜爱的压力，为了获得合适的算法反馈而进行优化。

Beyond the social media presences of individual people, businesses and cultural productions had to attract likes, too. A fashion brand might measure its success in Instagram likes, because more likes meant more customer engagement and potential buyers. Accounts made for new movies or television shows would likewise be oriented toward likes, because they were a measurement of marketing success: if you could get more followers for the account of the Twilight sequel New Moon, then more fans were excited for it and the debut would attract more viewers. Such is the pressure to be likable, optimizing for the appropriate algorithmic feedback.

这并不是什么新压力；渴望被喜欢是人类的基本心理，我们最终会喜欢与我们最相似的人，就像在谈话中模仿他人的肢体语言可以让你显得更有说服力或更具同理心一样。受欢迎程度将社会凝聚在一起，共同的动机是不冒犯或疏远他人。然而，人际受欢迎程度并非衡量文化的通用指标，尤其是在过去一两个世纪里我们珍视的创新艺术领域。艺术本身——更不用说作为人的艺术家——往往不受对受欢迎程度的追求所束缚，然而，在当今量化的暴政下，点赞才是最重要的。在“过滤世界”中，讨人喜欢的事物会成功，不讨人喜欢的事物注定会失败，尤其是在任何需要观众才能生存的文化领域。而且由于我们美国文化景观几乎完全服从于资本主义，这意味着或多或少全部服从于资本主义。

This is not a new pressure; it’s basic human psychology to want to be liked, and we ultimately like what is most similar to us, the same way that mirroring another person’s body language in conversation can make you appear more convincing or empathetic. Likability binds society together with the shared incentive to not offend or alienate. Yet interpersonal likability has not been a common metric of culture, particularly the kind of innovative art we have prized over the past century or two. Art itself—not to mention artists as people—tends not to be bound by the quest for likability, and yet likes are what the current tyranny of quantification prioritizes most. In Filterworld, what is likable succeeds and what is not likable is doomed to fail, particularly in any arena of culture where audiences are requisite for survival. And because our American cultural landscape is almost entirely subservient to capitalism, that means more or less all of it.

算法文化的空虚

THE EMPTINESS OF ALGORITHMIC CULTURE

2021年，马丁·斯科塞斯，这位在艺术和商业上都获得巨大成功的电影创作者，在《哈泼斯杂志》上发表了一篇文章。这本杂志与其说是世界著名导演的舞台，不如说是以刊登小众文学评论而闻名。在文章中，斯科塞斯表达了他对费里尼的个人欣赏，这位20世纪中叶的意大利导演，作品庞大，制作宏大，但他也借此机会批判了当代电影的现状。斯科塞斯写道，在流媒体视频时代，电影被扁平化地归入“内容”类别。“内容”成为所有动态影像的商业术语：一部大卫·里恩的电影、一段猫咪视频、一则超级碗广告、一部超级英雄续集、一部连续剧集。”斯科塞斯接着描述了我们文化生态系统的架构——我们消费的内容正在被算法推荐过滤，这些推荐基于我们已经看过的内容以及当前内容的主题或类型。他写道：“算法，顾名思义，就是基于将观看者视为消费者而非其他任何事物的计算。” 与内容互动的方式只有一种：吸收并点赞。

In 2021, Martin Scorsese, creator of films both artistically and commercially successful, published an essay in Harper’s, a magazine better known for hosting niche literary criticism than acting as a stage for world-famous directors. In the essay, Scorsese recounted his personal appreciation for Fellini, the mid-century Italian director of sprawling, grandiose productions, but he also took the opportunity to decry the state of contemporary cinema. In the era of streaming video, films had been flattened into the category of “content,” Scorsese wrote. Content “became a business term for all moving images: a David Lean movie, a cat video, a Super Bowl commercial, a superhero sequel, a series episode.” Scorsese goes on to describe the architecture of our cultural ecosystem—the content that we consume is being filtered by algorithmic recommendations, which operate based on what we’ve already seen and the subject matter or genre of the content at hand. “Algorithms, by definition, are based on calculations that treat the viewer as a consumer and nothing else,” he wrote. There is only one way to interact with content: ingest it and like it.

斯科塞斯认为，这种一刀切的内容分类和算法优先考虑熟悉度的做法已经损害了电影这种媒介的本质：“电影一直以来都远不止于内容，而且永远如此。” 被遗忘的是电影更深层次的艺术形式，这种改变了他以及许多人人生的媒介，以及通过银幕带来的美学甚至道德挑战。观看伟大的电影并不总是令人感到舒适的；这种体验超越了平庸的消费，旨在质疑社会规范，并帮助观众发现新的自我意识。

That blanket category of content and the algorithmic prioritization of familiarity has undermined the medium of film, Scorsese argued: “The cinema has always been much more than content, and it always will be.” What gets lost is the deeper art form of cinema, the medium that changed his life and the lives of so many others, the aesthetic and even moral challenges that come through the silver screen. Watching great movies was not always comfortable; the experience went beyond banal consumption and aspired to interrogating social norms and enabling viewers to discover new senses of self.

斯科塞斯以费里尼的电影为例，认为其内容并非易消化，而是电影制作艺术的巅峰。费里尼1963年的电影《八部半》是斯科塞斯导演作品的巅峰之作。这是一部关于艺术家人生的片段式、自我指涉式的沉思。斯科塞斯描述了这部电影首映时的情景：“人们为此争论不休：它的效果是如此戏剧化。我们每个人都有自己的解读，我们会整晚坐着讨论这部电影——每一个场景，每一秒。当然，我们从未最终确定一个明确的解读。”这部电影如此陌生，如此不为人知，以至于斯科塞斯不得不慢慢消化它，并将其影响融入到自己后来的电影中。我从斯科塞斯的文章中听到的偏执是，21世纪的艺术不再经得起这样的审视。相反，它廉价而短暂，在你的生活中飘荡，不会留下任何明显的痕迹。 （他写作的热情表明了斯科塞斯在多大程度上受到了费里尼的影响，六十年后他仍在承受这种影响。）这可能是因为，为了适应数字信息流，为了吸引那些有害的点赞并尽可能地进一步推销自己，文化必须首先是内容，其次才是艺术——如果有的话。

Scorsese used the example of Fellini’s films as the opposite of digestible content and the pinnacle of filmmaking as art. Fellini’s 1963 film 8½ is the pinnacle of the director’s oeuvre for Scorsese, a fragmentary, self-referential meditation on the life of an artist. Scorsese described the moment the movie debuted: “People argued over it endlessly: the effect was that dramatic. We each had our own interpretation, and we would sit up till all hours talking about the film—every scene, every second. Of course, we never settled on a definite interpretation.” It was so strange and unfamiliar that Scorsese had to digest it slowly, incorporating its influence into his own later movies. The paranoia that I hear in Scorsese’s essay is that the art of the twenty-first century no longer holds up to such scrutiny. Instead, it’s cheap and ephemeral, wafting through your life without leaving any discernible mark. (The passion of his writing shows just how much Scorsese was marked by Fellini, an impact that he was still processing six decades later.) That may be because to fit into digital feeds, in order to attract those pernicious likes and further promote itself as much as possible, culture has to be content first and art second—if at all.

斯科塞斯的抱怨可以归因于他作为电影界元老的地位；有些人甚至认为他的地位在倒退。世界自他年轻以来已经发生了变化；他不再需要思考什么是新的，因为他的声誉和接触面意味着他可以创作任何他想创作的东西。但也有人和他一样感到厌倦和焦虑，这与斯科塞斯的哀叹如出一辙，他哀叹算法推荐剥夺了文化的内在意义。制片人巴里·迪勒在接受美国国家公共电台（NPR）采访时评论道：“这些流媒体服务一直在制作他们所谓的‘电影’。它们不是电影。它们只是一些奇怪的算法过程，创造出了时长大约100分钟的东西。” 2021 年，评论家迪安·基西克 (Dean Kissick) 写道：“如今，很多文化都给人一种空洞、空洞的感觉，仿佛是由算法创造出来的。”他是当代文化领域最敏锐的评论家之一。“算法”已经成为任何感觉过于华而不实、过于简化或过于优化而无法吸引注意力的事物的代名词：高产值与对基本内容的漠不关心的结合。

Scorsese’s complaints can be chalked up to his position as an elder statesman of his medium; some may even find his position retrograde. The world has changed since his youth; he no longer needs to think about what is new, because his reputation and level of access mean he can create whatever he wants. But others shared his sense of ennui and anxiety, echoing Scorsese’s lament that something about algorithmic recommendations has robbed culture of its innate meaning. In an interview with NPR, the producer Barry Diller commented, “These streaming services have been making something they call ‘movies.’ They ain’t movies. They are some weird algorithmic processes that has created things that last 100 minutes or so.” “Much of culture now has the hollow, vacant feeling of having been made by algorithm,” wrote the critic Dean Kissick, one of the more incisive commentators on contemporary culture, in 2021. “Algorithmic” has become a byword for anything that feels too slick, too reductive, or too optimized for attracting attention: a combination of high production values with little concern for fundamental content.

我也感受到了那种空虚。到了2010年代末和2020年代初，许多文化形式——书籍、电视节目、电影、音乐和视觉艺术——似乎主要是为了吸引短暂的注意力，并填充无尽的应用程序信息流。没有什么能停留在杰作在几十年后仍会被重新审视，而这种保留下来的方式正是我所理解的。对我来说，一个重要的信号来自于艺术界流行的一种绘画风格，年轻艺术家们从画廊和拍卖会上赚得盆满钵满。2014 年，艺术评论家兼画家沃尔特·罗宾逊创造了“僵尸形式主义”一词来形容这种风格。僵尸形式主义是被剥夺了情感和宏伟的抽象表现主义，奥斯卡·穆里略和雅各布·卡塞等人的画布上充满了模糊的笔触或冷色调的单色。评论家杰里·萨尔茨也将其称为“雷同的艺术”。他们倾向于毫无意义的装饰，这导致了许多画家描绘光鲜亮丽的超现实主义场景，比如艾米丽·梅·史密斯的拟人化扫帚画作。（迪安·基西克称之为“僵尸形象”。）

I feel that emptiness, too. By the late 2010s and early 2020s, it seemed as though many cultural forms—books, TV shows, movies, music, and visual art—primarily existed to garner ephemeral attention and populate the endless app feeds. Nothing stuck in the way that a masterpiece that will be revisited decades hence sticks. For me, a major signal came from the style of painting that became popular in the art world, with young artists netting huge prices from galleries and auctions. In 2014, the art critic and painter Walter Robinson coined the term “zombie formalism” to describe it. Zombie formalism was abstract expressionism shorn of its emotion and grandeur, with canvases of mushy brushstrokes or cold monochromes from the likes of Oscar Murillo and Jacob Kassay. The critic Jerry Saltz echoed that it was “look-alike art.” Their tendency toward the meaninglessly decorative led the way to a slew of painters depicting glossy surrealist scenes, like Emily Mae Smith’s paintings of anthropomorphized broomsticks. (Dean Kissick labeled it “zombie figuration.”)

这些画作是为Instagram改编的，收藏家们也越来越倾向于在Instagram上发现并购买这些作品（通常无需亲眼看到作品），而这同样是通过算法推荐进行中介的。他们可以像转发帖子一样轻松地在同一个平台上转售这些作品。2014年，臭名昭著的挪用艺术家理查德·普林斯（Richard Prince）简化了这一流程，创作了一系列“画作”，实际上是从Instagram帖子中发现的印刷品复制品，售价高达10万美元。

These were paintings adapted for Instagram, which was also where collectors increasingly tended to discover and buy them (often without seeing the art in person), mediated yet again by algorithmic recommendations. They could resell it frictionlessly on the same platforms as easily as resharing a post. In 2014, the infamous appropriation artist Richard Prince shortcut the process and produced a series of “paintings” that were actually printed replicas of found Instagram posts, which sold for prices up to $100,000.

对算法驱动艺术的恐惧，部分源于艺术家的缺失：如果可行的艺术作品可以由计算机创作或策划，那么人类创作的意义何在？像迈克·温克尔曼（Mike Winkelmann，艺名 Beeple）这样的艺术家，凭借其草率的 CGI 动画在 Instagram 上积累了超过 200 万粉丝，但这种成功需要他每天更新作品，而且他的智力水平很少超过一个十三岁男孩。（在 2021 年非同质化代币热潮初期，Beeple 的艺术作品在佳士得拍卖行以 6900 万美元的价格成交，这既是玩笑，也是对其受欢迎程度的认真展现。）作为一名创作者，人们很容易担心自己即将被淘汰。但消费者也同样受到影响。

Part of the fear of algorithmically driven art is the obviation of the artist: If viable art can be created or curated by computer, what is the point of the humans producing it? An artist like Mike Winkelmann, known by his alias Beeple, built up over two million followers on Instagram with his slapdash CGI cartoons, but that success required posting every day and rarely rising above the intellectual level of a thirteen-year-old boy. (Early in the 2021 boom for nonfungible tokens, Beeple artwork sold at Christie’s for $69 million, both a joke and an entirely serious demonstration of popularity.) It’s easy to worry about your own looming obsolescence as a creator. But consumers are equally affected.

音乐杂志《Pitchfork》的编辑杰里米·拉尔森（Jeremy Larson）在2022年的一篇文章中抱怨，Spotify 的算法听歌体验妨碍了音乐本身。“尽管它包含了我想要的所有音乐，但这些音乐都缺乏成就感、情感共鸣或个人化，”拉尔森写道。尽管艺术家的意图或许并未改变，“音乐成为了流媒体服务的广告，你投入的时间和注意力越多，科技公司就越受益。” 平台成为了听众与艺术家及其作品之间关系的过滤器，有时甚至是障碍。并非所有艺术家都允许流媒体服务播放他们的音乐。像尼尔·杨和琼尼·米切尔这样的音乐家已经从Spotify下架了他们的曲目，以抗议该公司的一些决定，包括资助那些宣扬政治和文化阴谋论的播客。艾丽娅的音乐目录的数字版权谈判耗时数年；直到2021年，这些曲目才在Spotify上发布。这些音乐家的缺席使得他们更难接触到，Spotify的订阅用户或许更容易忘记——或者从一开始就不会发现——因为这项服务涵盖了用户的大部分收听习惯。即使听众可能喜欢艾丽娅、杨或米切尔，Spotify的推荐算法也无法推荐他们，因为Spotify无法从这些收听中获利。它被动地限制了我们感知音乐的方式。拉尔森将其描述为“一种虚构的现实，旨在取代应用程序之外随机的生活轮廓。”

In a 2022 essay, Jeremy Larson, an editor at the music magazine Pitchfork, complained that the algorithmic experience of listening to music on Spotify was getting in the way of the music itself. “Even though it has all the music I’ve ever wanted, none of it feels necessarily rewarding, emotional, or personal,” Larson wrote. Though the artists’ intentions may not have changed, “music becomes an advertisement for the streaming service, and the more time and attention you give it, the more it benefits the tech company.” The platform becomes a filter, and sometimes a barrier, for the listener’s relationship to the artist and their work. And not every artist allows the streaming service to broadcast their music in the first place. Musicians like Neil Young and Joni Mitchell have removed their catalogs from Spotify to protest some of the company’s decisions, including funding podcasts that promulgate political and cultural conspiracy theories. It took years to negotiate digital rights for Aaliyah’s catalog of music; it wasn’t available on Spotify until 2021. The musicians’ absence makes them harder to access and perhaps easier for a Spotify subscriber to forget about—or never discover in the first place—since the service encompasses so much of its users’ listening habits. Its recommendation algorithms would fail to promote Aaliyah, Young, or Mitchell even if a listener might like them because Spotify can’t profit from the listen. It passively limits how we perceive music. Larson described it as “a fabricated reality meant to replace the random contours of life outside the app.”

拉森用一个引人入胜的比喻来比喻我们集体的算法消费：“如今，数百万用户并肩坐在一个巨大的录制音乐槽边，每月只需支付一份 Chipotle 墨西哥卷饼的价格。” 这既是象征性地，也是字面意义上的媒介贬值；尽管订阅价格低廉，但 Spotify 几乎没有将其中的一小部分转嫁给真正的艺术家。在自助式流媒体服务和无限量推送出现之前，黑胶唱片、磁带或 CD 的稀缺性促使听众去了解艺术家的作品，因为否则，在音乐上的金钱投入可能会感觉不值得。算法推送的承诺是，如果音乐变得无聊或令人厌倦，你可以换到下一首歌。下一首推荐的歌曲可能会符合你预先设定的品味，而你无需为此支付任何额外费用。

In an evocative metaphor for our collective algorithmic consumption, Larson wrote that “millions of users now sit side by side at the ledge of one great big trough of recorded music for the monthly price of a Chipotle burrito.” It’s a symbolic and literal devaluing of the medium; as low as that subscription price is, Spotify passes on little of it to actual artists. Before the advent of buffet-style streaming services and infinite feeds, the scarcity of a single vinyl record, cassette tape, or CD provided an incentive for a listener to get to know an artist’s work, because otherwise the monetary investment in the music might not feel worth it. The promise of an algorithmic feed is that if the music becomes at all boring or tiresome, you can flip to the next song. That next recommendation will probably stick to the limits of your predetermined taste, and you won’t have paid anything more for it.

我在观看Netflix剧集时，尤其是一次性狂看多集剧集时，也会有和拉尔森描述的一样的感觉。当然，这些剧集确实很精彩——精彩到让我停不下来。但我却说不出有多少Netflix制作的剧集能做到一直萦绕在我的心头。我个人对美食纪录片的流媒体痴迷。这类节目我都会看：比如介绍各大洲街头美食的旅行节目，或者讲述米其林星级厨师的传记片，片中充满了牛排上烤的B卷。鉴于这类节目很少有魅力十足的主持人——他们的出席会让制作大量剧集变得更加困难——它们在我的记忆中都像一个长长的屏幕保护程序一样模糊不清。这些节目相当于旅游景点千篇一律的Instagram帖子——充斥着空洞的点赞和毫无意义的互动，无休止地复制内容。

I get the same feeling that Larson describes when I watch Netflix shows, especially when bingeing multiple episodes of a series at once. Sure, the shows are enjoyable—so enjoyable that I can’t stop watching them. But I can’t name many Netflix-produced shows that have stuck with me. My personal streaming addiction is to food documentaries. I’ll watch anything within that category: travel shows highlighting street food from different continents or a biopic that follows a Michelin-starred chef with copious B-roll of steaks hitting the grill. Given that few of these productions have charismatic individual hosts—their required presence would make it more difficult to manufacture episodes in high volume—they all blur together in my memory like one long screen saver. These shows are the equivalent of monotonous Instagram posts from tourist destinations—fodder for empty likes and thoughtless engagement, the endless reproduction of content.

这并不是说这样的内容不能充满艺术感；Netflix 收购的《寿司之神》及其后续剧集《主厨的餐桌》等作品的创新之处在于，它们将视觉美感放在首位，并使用柔焦镜头拍摄食物特写。他们将 Instagram 上的美食照片搬上了电视荧幕。然而，这种艺术感与平静、低调和氛围的需求紧密相连——打造出堪比完美亚麻床单的视听效果。这些作品不激发任何思考，只带来感官享受。与艺术影院或独立录像带租赁店为确保每部作品独具特色而进行的精心策划不同，这些节目无需提供意义非凡的一次性体验；它们可以作为一种麻醉剂大量存在。

It’s not that such content can’t be artful; the innovation of productions that Netflix acquired like Jiro Dreams of Sushi and the follow-up series Chef’s Table was to focus on visual beauty above all, with soft-focus cameras trained on food close-ups. They translated the food porn of Instagram photos into television. Yet the artfulness was yoked to the need to be placid, undisruptive, and ambient—developing the audiovisual equivalent of perfect linen bedsheets. The productions inspire no thoughts, only sensory pleasure. Unlike the curation required in an art-house cinema or an independent video rental store to ensure that each offering stands on its own, the shows didn’t have to provide meaningful one-off experiences; they could simply exist in bulk as a numbing narcotic.

与氛围美食纪录片的空洞热量相比，美食频道（Food Network）在20世纪90年代和21世纪初的兴起，推动了名厨的演变，并引发了家庭烹饪文化的巨变。在流媒体服务中，真正的烹饪节目明显缺席，仿佛它们会鼓励过多的体力活动，从而减少人们看屏幕的时间。剩下的只是纯粹的、毫无成效的、催眠式的娱乐，因为其核心目的仅仅是让观众以“活跃用户”的身份离开服务。Netflix甚至制作了其节目的复制品，故事发生在不同的国家，使用不同的语言。《回家过圣诞节》最初是一部挪威迷你剧，讲述了一位单身女性在乡村小镇勒罗斯试图在节日前找到男朋友的故事；它几乎被翻拍成意大利电影《我讨厌圣诞节》，故事发生在基奥贾——这是一种廉价的……内容翻倍。一旦某个公式奏效，它就可以在 Netflix 庞大的全球观众群中复制或扩展，最终导致他们不知不觉地消费了相同的内容。复制的节目可以通过算法推荐提供给任何可能感兴趣的观众。

Contrast the empty calories of ambient food documentaries with the rise of the Food Network cable channel in the 1990s and 2000s, which prompted the evolution of chef-celebrities and caused a sea change in the culture of home cooking. On streaming services, actual cooking shows are notably absent, as if they might encourage too much physical activity that would take away from time spent looking at a screen. What remains is pure, unproductive, hypnotic entertainment, because the core purpose is simply to have the viewers leave the service on as “active users.” Netflix has gone so far as to produce replicas of its shows, set in different countries and using different languages. Home for Christmas began as a Norwegian miniseries about a single woman in the rural town of Røros trying to find a boyfriend before the holidays; it was remade almost shot by shot in the Italian I Hate Christmas, set in Chioggia—a cheap way to double your content. Once a formula works, it is repeatable, or scalable, across Netflix’s vast global audience, who end up unknowingly consuming the same material. The replicated show can be served to any possibly interested viewer via algorithmic recommendation.

在流媒体时代早期，Netflix 因其自动播放功能而臭名昭著，该功能于 2016 年首次推出。当一集电视剧或电影结束时，计时器会倒计时十秒，然后开始播放另一部电视剧或电影，要么是该系列的后续剧集，要么是算法推荐的替代剧集。在 2019 年 Hacker News 论坛的一篇帖子中，一位 Netflix 工程师回忆说，最初的十秒是“观看时长增幅最大的原因”；五秒太过刺耳，十五秒又太慢。当时，自动播放功能感觉像是一种彻底的背离。电视剧不是应该直接停止播放吗？在有线电视上，你通常要等一周才能看到下一集。但由于 Netflix 通常会一次性添加整季电视剧，这项功能几乎强制用户狂看。它还会促使用户沿着算法的消费路线前进，可能会坚持观看某一特定类型，比如简·奥斯汀翻拍的作品或外星人动作片，因为它会一个接一个地推荐。 （对于推荐算法来说，多样性是一个困难的概念。）

Earlier in the streaming era, Netflix was infamous for its autoplay feature, which was introduced in its first form in 2016. When a TV episode or movie ended, a timer counted down ten seconds and then another show or film started, either the subsequent episode in the series or an algorithmically recommended alternative. In a 2019 post on the forum Hacker News, one Netflix engineer recalled that the original ten seconds caused “the biggest increase in hours watched”; five was too jarring and fifteen too slow. At the time, autoplay felt like a radical departure. Wasn’t a TV episode simply supposed to stop? On cable, you’d usually have to wait a week for the next episode in a series. But since Netflix most often added entire television seasons at once, the feature all but mandated binge-watching. It also pushed users to coast along algorithmic lines of consumption, perhaps sticking to a given genre, like Jane Austen remakes or action movies featuring aliens, as it recommended one example after another. (Variety is a difficult concept for recommendation algorithms.)

如果Netflix连续播放了三集而用户没有停止自动播放功能，或者连续观看了90分钟后，应用甚至会自动停止，屏幕上会弹出一条重要的信息：“你还在看吗？”这个功能一直延续至今。某种程度上，它是为了防止用户在电视机前睡着后服务仍然保持开启状态。每当我遇到这种情况，无论是在夜晚昏暗的客厅里，还是从我的笔记本电脑屏幕上发出，我都会感到一丝尴尬。不是因为我睡着了，而是因为我看的电视比平时多得多，一集接一集地滑下去，我都来不及停下来。平台本身鼓励我沉迷其中，但警告信息却暗示这样做是有害的。然而，到了2020年代，自动播放已成为YouTube和TikTok的常态：你永远不会指望一个信息流会结束。如今，所有文化都是内容，而我们用来访问它的平台鼓励我们将其视为可以互换的。

After Netflix played three episodes in a row without the user stopping the autoplay function, or after ninety minutes of continuous watching, the app even stopped itself and a fateful message appeared on screen asking, “Are you still watching?” This feature persists today. In part, it was a safeguard against the service remaining on after the user fell asleep in front of the TV. The times that I encountered it, in the dim illumination of the living room at night or emanating from my laptop screen, I felt a tinge of embarrassment. It wasn’t that I was asleep, it was just that I was watching way more TV than I usually did, one episode sliding implacably into the next before I had the willpower to stop it. Bingeing was something that the platform itself encouraged me to do, and yet the warning message implied that it was negative. By the 2020s, however, autoplay had become the norm via YouTube and TikTok: you would never expect a feed to end. All culture is now content, and the platforms we use to access it encourage us to treat it as interchangeable.

2007 年，亚马逊推出了 Kindle Direct Publishing，电子书市场，这些电子书在其新款 Kindle 电子阅读器上阅读。随着时间的推移，KDP 成为自出版作家的聚集地，他们通过线上平台绕过了传统的代理商、编辑和书店层级结构。新书会自动推荐给读者，就像亚马逊商店里推荐衬衫或搅拌机一样。KDP 上成功的内容也与传统文学界的成功截然不同。这是一个以文学为内容的空间，主题的特异性和字数远远超过了评论家的意见。作者创作的书越多越好。据估计，KDP 在 2022 年提供了超过 1200 万本电子书。亚马逊不仅控制着电子书；2019 年，其规模更大的数字店面占据了成人新书在线销售额的四分之三，几乎占了所有新书销售额的一半。换句话说，许多文学作品被迫通过亚马逊平台到达消费者手中，这迫使书籍形成特定的形式——局限于特定类型类别的大批量系列，并随着时间的推移以少量内容的形式持续发布——就像推文必须以特定形式发布才能在 Twitter 推送中取得成功一样。

In 2007, Amazon launched Kindle Direct Publishing, a marketplace for eBooks, which were consumed on its new Kindle e-reader. Over time, KDP became a hub for self-published authors who circumvented the traditional hierarchy of agents, editors, and bookstores by going online, where new books were automatically recommended to readers using the same mechanisms as blouses or blenders in the Amazon store. The content that succeeded on KDP was a departure from what succeeded in the established literary world, too. It was a space for literature-as-content, where subject specificity and word volume far outweighed the opinions of critics. The more books an author produced, the better. According to one estimate, KDP offered over twelve million e-books in 2022. Amazon wasn’t just controlling e-books; in 2019, its larger digital storefront made up three-quarters of online sales of new books for adults and almost half of all new book sales overall. In other words, much of literature is forced to move through Amazon’s platform to reach consumers, which pressures books into particular forms—high-volume series confined to specific genre categories and released consistently over time in a drip of content—the same way a tweet has to be written to succeed in the Twitter feed.

斯坦福大学教授马克·麦格尔研究了现代主义文学在二十世纪的演变，包括追踪小说艺术硕士（MFA）课程的发展如何影响二十世纪中期的小说。小说家出身的教授，为了维持写作，在爱荷华作家工作室等机构指导学生，经常从个人视角引导他们走向一种自觉的文学现实主义风格。温德尔·贝里、理查德·福特、迈克尔·夏邦、里克·穆迪和塔玛·贾诺维茨都是MFA课程模式的成功范例。如今，MFA课程仍然扮演着守门人的角色，帮助出版商发现新人才，并将小说家引入职业化行业。这类课程的精挑细选，以及出版业封闭的、“握手”式的运作方式，使得它们能够推广独特或富有挑战性的艺术家——这是一种品味创造的行为——尽管它们本质上带有某种精英主义色彩。

The Stanford University professor Mark McGurl has studied how modernist literature evolved through the twentieth century, including tracking how the development of fiction MFA programs influenced mid-century novels. Novelists-turned-professors, taking jobs to support their writing practice, tutored their students at institutions like the Iowa Writers’ Workshop, often pushing them toward a style of self-conscious literary realism from a personal perspective. Wendell Berry, Richard Ford, Michael Chabon, Rick Moody, and Tama Janowitz were some of the successes of the MFA-program model. Today, MFA programs still act as gatekeepers, helping publishers identify new talent and ushering novelists into the professionalized industry. The handpicked nature of such programs, and the insular, handshake nature of the publishing industry, maintain the ability to promote a singular or challenging artist—an act of tastemaking—though they are by nature somewhat elitist.

麦格尔指出了艺术硕士（MFA）风格即将趋于同质化。但算法守门人可以先行。年轻作家往往在进入网络之前就想方设法培养公众影响力。在推特、Instagram 或抖音上，艺术硕士项目（MFA）项目。他们将自己的声音置于社交媒体扁平化的力量之下。这些预先准备好的角色甚至可能在竞争激烈的研究生申请中有所帮助。在这个过程的每一步，文学创意都会接受测试，看其能否在关注的市场中吸引在线互动。

McGurl identified the looming homogeneity of MFA style. But algorithmic gatekeepers can come first. Young writers often find ways to cultivate public presences online even before they enter MFA programs, on Twitter, Instagram, or TikTok. They subject their voices to the force of social media flattening. These pre-prepared personae might even help in competitive grad school applications. At each step of the process, a literary idea is tested for its ability to draw online engagement in a marketplace of attention.

麦格尔如今认为，我们正进入文学界的亚马逊时代，在这个时代，亚马逊既是美学的仲裁者，也是商业的仲裁者。麦格尔在其2021年出版的《一切与更少：亚马逊时代的小说》一书中写道，它“正在将自己打造为文学生活的新平台”。该平台衡量质量的指标是数量，与其他算法推送一样，也是衡量参与度的无情标准。购买量越多，阅读页数越多，就意味着一本书比同类书籍更优秀。书籍封面不仅要设计得在小屏幕上清晰易读，而且行文也必须针对翻页进行优化——用每一行文字抓住读者的注意力。（当然，优秀作品通常都具备这种特质，但并非总是如此。）

McGurl now sees us entering into an Amazon era of literature, in which the company is an aesthetic as well as a commercial arbiter. It is “offering itself as the new platform of literary life,” McGurl wrote in his 2021 book Everything and Less: The Novel in the Age of Amazon. The platform’s measure of quality is quantity, the same ruthless metric of engagement as other algorithmic feeds. More purchases, and more pages read, meant that a book was better than its peers. Not only did a book cover have to be designed to be legible on a small screen, but the writing also had to be optimized for page turning, as it were—grabbing the reader’s attention with each successive line. (This is a quality that good writing often has, of course, but not always.)

一方面，这是一种民主化：任何人都可以出版书籍，并有机会通过完全相同的渠道以相同的方式销售。没有书店购书员或前台策划的障碍；只有算法的计算。超级畅销书作家科琳·胡佛就是一个这样的例子。胡佛最初在亚马逊上自费出版她的小说，这些小说通常属于浪漫、惊悚和青少年小说类别。她的前两部小说大获成功，促使主流出版商Atria Books在2012年重新出版了它们，她的小说开始持续登上畅销书排行榜。（尽管有其他选择，胡佛仍然坚持自费出版她的第三部小说。）疫情期间，胡佛免费提供她旧作的电子书——这是另一种提升数字互动的策略——并在TikTok上吸引了大批粉丝，新兴的图书社区纷纷为她作品的影响力而感动落泪。据《纽约时报》报道，她二十多本书的销量估计超过两千万册，“比詹姆斯·帕特森和约翰·格里森加起来还要多” 。飞轮甚至在胡佛的控制之外也加速运转：“我读了别人的书，真是羡慕不已。我在想，天哪，这些“品质好多了，为什么我的产品卖得那么差？” 2022 年，她告诉《纽约时报》 。

On one hand, this is a kind of democratization: Anyone can publish a book and give it a chance to be sold through the exact same channels, presented in the same way. There is no obstacle of a store’s book buyer or the curation of a front table; just the math of the algorithm. The hyper-bestselling author Colleen Hoover provides an example of the opportunities. Hoover began by self-publishing her novels, which often fall into romance, thriller, and young-adult categories, on Amazon. The success of her first two led a mainstream publisher, Atria Books, to republish them in 2012, and her novels began to consistently hit bestseller lists. (Hoover stuck with self-publishing for her third novel, despite other offers.) During the pandemic, Hoover made eBooks of her backlist free—another strategy to boost digital engagement—and sparked a wave of fans on TikTok, where the nascent book community made tearful testaments to her writing’s impact. Sales of her more than twenty books are estimated above twenty million copies, “more books than James Patterson and John Grisham combined,” as The New York Times reported. The flywheel has accelerated even outside of Hoover’s control: “I read other people’s books, and I’m so envious. I’m thinking, Oh my God, these are so much better, why are mine selling the way they are?” she told the Times in 2022.

另一方面，大众参与的要求背离了文学史，在文学史上，编辑和学者的意见远比一本书最初的销量重要得多。麦格尔写道，文学经典“与亚马逊毫无关联，除了将其作为学生倾向于购买的书籍清单”。他写道，经该平台筛选，“所有小说都是类型小说”，无论是耗时十年创作的实验小说，还是情色电子书系列的第五卷。麦格尔指出，某些类型在亚马逊的KDP市场中获得了成功。言情小说表现不俗，包括“阿尔法亿万富翁言情小说”（最著名的例子是《五十度灰》）和“三人行MMF军人”等小众子类型。（这些类型名称本身与搜索引擎优化语言相似并非巧合。）史诗小说也大受欢迎，其中包括庞大的奇幻系列。

On the other hand, the requirement of mass engagement is a departure from the history of literature, in which the opinions of editors and academics have mattered far more than how many copies of a book initially sells. The literary canon, McGurl wrote, is “a thing Amazon has no particular relation to at all except as a list of books that students tend to purchase.” As filtered by the platform, “all fiction is genre fiction,” he wrote, whether an experimental novel that took a decade to write or the fifth volume in an endless series of erotica eBooks. McGurl identified certain genres that found a successful home in Amazon’s KDP marketplace. Romance books played well, including niche subgenres like “alpha billionaire romance,” Fifty Shades of Grey being the most famous example, and “threesome MMF military.” (It’s not a coincidence the genre names themselves resemble search-engine optimization language.) Epics were another hit, including sprawling fantasy series.

亚马逊文学现象最容易被观察到的地方，仅限于一个相对狭小的空间：数字书店和 Kindle 阅读器。两者都局限于特定场景，或许尤其适合阅读“罪恶快感”之类的活动——在 Kindle 上，没有人能看到你正在阅读的书名。避开算法的影响，去实体书店也很容易，店员可能会进行个性化推荐。然而，麦格尔也将亚马逊塑造的读者消费习惯与 2010 年代自传体小说等高阶文学类型的兴起联系起来——这些出版业的梯队在当时并不被视为算法驱动或市场驱动的领域。

At its most observable, the phenomenon of Amazon Literature is confined to a relatively small space: the digital storefront and the Kindle reader. Both are relegated to specific situations and might be particularly amenable to activities like reading guilty pleasures—on Kindle, no one can see the title of the book you’re reading. It’s easy enough to avoid the algorithmic influence and go to a physical bookstore, where a clerk might make a personal recommendation. Yet McGurl also linked the consumption habits that Amazon conditioned readers into with the rise of high-literary genres like 2010s autofiction—echelons of the publishing industry that were not seen as so algorithmic nor market-driven.

麦克格尔写道，自传体小说是一种“以一个几乎不具象化的作家主角为中心”的体裁。尽管它起源于20世纪70年代的法国，由理论家塞尔日·杜布罗斯基（Serge Doubrosky）首创，但近年来随着希拉·赫蒂（Sheila Heti）、本·勒纳（Ben Lerner）、雷切尔·卡斯克（Rachel Cusk）和卡尔·奥韦·克瑙斯加德（Karl Ove Knausgaard）等小说家的成功而开始流行。他们的作品通过作者与叙述者之间密切而又模糊的关系联系在一起：卡斯克2014年开始的《大纲》三部曲中的“我”究竟是卡斯克本人，就像一本回忆录，还​​是……叙述者与纯小说中的事件之间有什么关联？其吸引力在于那种偷窥式的张力，让人猜不透哪个是哪个。当然，读者对社交媒体上的这种动态非常熟悉，无论是在推文还是照片中，其他人都会以不同程度的真实性展现自己的生活和自我。自传体小说有点像网红的Instagram账号：碎片化、缺乏叙事性，而且常常带有欺骗性。

Autofiction is a genre “centering on a barely fictionalized writer-protagonist,” McGurl wrote. Though it originated in France in the 1970s, coined by the theorist Serge Doubrosky, it came into vogue more recently with the success of novelists including Sheila Heti, Ben Lerner, Rachel Cusk, and Karl Ove Knausgaard. Their work is connected by a close but ambiguous relationship between author and narrator: Is the “I” of Cusk’s Outline trilogy of novels beginning in 2014 really Cusk herself, like a memoir, or are the narrator and the events within pure fiction? The appeal comes from the voyeuristic tension of guessing which is which. Of course, readers are intimately familiar with this dynamic from social media, where other people present their lives and selves with varying degrees of truthfulness, whether in tweets or photos. Autofiction is a bit like an influencer’s Instagram account: fragmented, non-narrative, and often deceptive.

根据麦格尔的分析，卡尔·奥韦·克瑙斯加德和雷切尔·卡斯克的自传体小说都是系列作品，内容丰富，并以近乎消费主义的视角，展现了一位成功作家的人生。有时，这些书会逐渐演变成一种愿望的满足，至少对其文化精英目标读者而言是如此。人们购买这些书，就像购买远在异乡的住所和作家座谈会的生活方式，阅读它们就像观看真人秀节目。扎迪·史密斯曾这样评价《我的奋斗》 ：“我迫切需要下一卷。”这句话或许和你对《比佛利娇妻》一季的评价一样。我认为卡斯克和克瑙斯加德是21世纪两位颇具趣味的小说家，但我可能无意中忽略了他们看似前卫的文学风格在多大程度上掩盖了他们作品中更为平庸和主流的内容。最近重读时，我不禁想起卡斯克的《过境》（她的“大纲”三部曲中的第二部）里，有多少内容是关于沙龙理发和房屋装修的趣闻轶事。卡斯克是在彻底改变我与叙事的关系，还是我只是希望在伦敦一个理想的社区里有一套公寓，可以重新装修？

Per McGurl’s analysis, both Karl Ove Knausgaard and Rachel Cusk’s autofiction novels came in series, offering readers a high volume of content, and presented a vicarious, almost consumerist view of the life of a successful writer. At times the books shade into a form of wish fulfillment, at least for their cultural-elite target audience. One buys the book as if one is buying the lifestyle of far-flung residencies and author panels, reading it as if watching a reality TV show. “I need the next volume like crack,” Zadie Smith once wrote of My Struggle, the same thing you might say of a season of Real Housewives. I think Cusk and Knausgaard are two of the more interesting novelists of the twenty-first century, but I might unwittingly overlook the degree to which the seemingly avant-garde literary style masks their books’ much more banal and mainstream content. Upon a recent reread, I was reminded of just how much of Cusk’s Transit, the second in her Outline trilogy, consists of anecdotes about salon haircuts and home renovation. Is Cusk radically overhauling my relationship to narrative, or do I just wish I had a London flat to rebuild in a desirable neighborhood?

这里的重点并非克瑙斯高在描写自己在挪威乡村的青春岁月时关注的是Instagram的点赞数。而是算法塑造了整体的文化格局，影响了我们的品味。一切都存在于被动、无摩擦消费的算法语境中。即使一本书或其他内容似乎存在于算法生态系统之外，它仍然受到算法信息流所催生的主流美学和潮流的影响。算法文化的最终目标是持续不断地涌现出既相似又不同的内容，这些内容足够多样化，不会让人感到完全无聊，但也不会过于扰乱以至于让人感到疏远。追求一个雄心勃勃的艺术理想或许逐渐淡化，转而追求精细化，以实现点赞和参与为首要目标。

The point here is not that Knausgaard was paying attention to Instagram likes when he wrote about his youth in rural Norway. It’s that algorithms have shaped the overall cultural landscape, conditioning our tastes. Everything exists within the algorithmic context of passive, frictionless consumption. No matter that a book or other piece of content seems to exist outside of the algorithmic ecosystem; it is still informed by the dominant aesthetics and trends that algorithmic feeds have given rise to. The end point of algorithmic culture is a constant flow of similar-yet-different content, varied enough so as not to be utterly boring but never disruptive enough to be alienating. Reaching toward an ambitious artistic ideal may have faded in favor of refinement toward the goal of likes and engagement above all.

二十一世纪初，许多流行文化形式似乎沦为麻醉性的情绪提升，或是简单的谜题，任由观众自行解开谜题，然后转向下一个。即便是规模宏大的制作，亦是如此。像2019年的《复仇者联盟：终局之战》这样的电影，本应为漫威多年来的超级英雄电影画上句号，就像《星球大战》原版结局曾一度占据公众的想象力一样。然而，《终局之战》长达三个小时，比一般的大片长得多，它更注重特效和粉丝的冷知识——你可以在他们回归与反派进行最后一次战斗时认出你最喜欢的超级英雄——而不是令人满意的故事情节。漫威的粉丝们或许会感到欣慰—— “粉丝服务”一词指的是那些公然迎合铁杆粉丝需求的内容——但最终的成品在情感冲击或创意表达方面几乎毫无意义。

Many popular cultural forms in the early twenty-first century seem to have been reduced to either narcotic mood enhancements or simplistic puzzles left for the viewer to solve and then move on to the next. This extends to even our largest productions. A film like the 2019 Avengers: Endgame was supposed to provide a capstone to many years of Marvel superhero films, the way that the original Star Wars ending once occupied the public imagination. Across its three-hour run time, much longer than the usual blockbuster, Endgame prioritized special effects and checking the boxes of fan trivia—you can spot your favorite superhero as they all return to fight the villain for the last time—over satisfying storytelling. Marvel devotees may have been pleased—fanservice is the term of art for content that overtly caters to hardcore fans’ desires—but the end result is close to meaningless in terms of emotional impact or creative expression.

算法推送非但没有鼓励原创艺术成就，反而催生了对现有内容的需求，这些内容是为了生成更多内容：电影提供了现成的高潮场景GIF，可以在Twitter或TikTok上分享，还有那些能激发表情包的俏皮话，可以用来营销。这种对互动的需求可能会促使人们屈服于粉丝服务，或者至少是尝试这样做。

Rather than encouraging original artistic achievement, algorithmic feeds create the need for content that exists to generate more content: films that provide ready-made GIFs of climactic scenes to share on Twitter or TikTok and quippy lines that will inspire memes to serve as marketing. The need for engagement can encourage a capitulation to fanservice, or at least an attempt to do so.

2019年《权力的游戏》电视剧完结时也遭遇了同样的问题。两千万人观看了大结局——这是史上收视率最高的电视剧集之一。然而，许多观众却对结局感到失望。经过多季精心塑造的人物弧线，随着丹妮莉丝·坦格利安变得邪恶凶残，最终几集似乎将既定的人物形象抛诸脑后。剧中讲述的王室斗争演变成奇幻的战争，巨龙焚毁电脑特效城市，场面凌驾于剧情之上。在推特上分享的片段中，这或许看起来不错，但实际观看起来却毫无意义。由于小说家乔治·RR·马丁未能及时完成原著，剧集制作人大卫·贝尼奥夫和DB·韦斯不得不亲自策划大结局，这或许是数字化的又一次胜利。流媒体胜过文学。他们缺乏内在的远见，只追逐那些可能在网上获得良好反响的内容，并为这个复杂的谜题提供了简洁的答案。对于像马丁这样独特的叙事，优化并不奏效；尽管投入了巨额资金，但最终季却成为了昙花一现的内容，似乎一夜之间就从观众的脑海中消失了。

The original Game of Thrones TV series fell prey to the same problem when it ended in 2019. Twenty million people watched the finale—one of the most-watched TV episodes ever. And yet it left many viewers cold. After many seasons of carefully developed character arcs, the final episodes seemed to throw established personalities out the window as Daenerys Targaryen turned evil and murderous. The royal maneuvering that the show recounted devolved into fantastical battles and dragons burning down CGI cities, spectacle surpassing plot. It may have looked good in clips shared on Twitter but it was nonsensical to watch. It’s telling that the showrunners David Benioff and D. B. Weiss had to plot the finale themselves, since the novelist George R. R. Martin didn’t finish the book series in time—perhaps another triumph of digital streaming over literature. Lacking an internal vision, they chased what might play well online and provided tidy answers to the sprawling puzzle. Optimization didn’t work for a narrative so individual as Martin’s; despite the great expense that went into it, the final season became ephemeral content and faded out of viewers’ minds seemingly overnight.

爱尔兰作家莎莉·鲁尼的前三部小说被誉为西方千禧年小说的巅峰之作，是一部浪漫爱情三部曲，讲述了在爱尔兰和欧洲大陆阴郁的景色中成长的人物。这些作品氛围浓厚，引人入胜，充满了当地细节，并以鲁尼简洁、优雅却略带冷峻的笔触，营造出一种舒缓的氛围。小说见证了人物坠入爱河又失恋的历程，而爱情的美好往往发生在那些刺穿年轻自我自恋的罕见时刻。小说中也大量运用即时通讯和电子邮件记录，这是鲁尼笔下人物的原生沟通媒介。如此精准地反映我们的数字社交生活是她小说的一大亮点。除了间接的娱乐价值外，这些小说还描绘了各种社会问题，这些问题在小说出版时就成为了推特上的热门话题，评论家们在他们的评论文章中对相关话题发表了各种不同的看法（事实上，鲁尼学生时代就是辩论俱乐部的明星）。在《与朋友对话》中，它展现的是多角恋和自残。在《普通人》中，它展现的是受虐狂的性行为。在《美丽的世界，你在哪里》中，它展现的是经济阶层差异和文学名望本身，并逐渐走向自传体小说。

The first three novels by the Irish author Sally Rooney, currently upheld as a peak of Western millennial fiction, are a trilogy of romances that dramatize coming-of-age in moody Irish and continental European landscapes. They are atmospheric books, absorbing in their wealth of local detail and soothingly written in Rooney’s simple, elegant, and yet somewhat cold prose. The novels witness their characters falling into and out of love, love that happens best in rare moments when it pierces the narcissism of the young self. They also heavily feature instant-message and email transcripts, the native communication medium of Rooney’s characters. Reflecting our digital social lives so accurately is a strength of her novels. Alongside their vicarious entertainment value, they depict various social problems that became tentpoles of Twitter discourse when the books debuted, with pundits taking various sides of the topic at hand in their think pieces (in fact, Rooney was a debate-club star as a student herself). In Conversations with Friends, it was polyamory and self-harm. In Normal People, it was masochistic sexuality. In Beautiful World, Where Are You, it was economic class differences and literary fame itself, moving toward autofiction.

这三部小说都引发了关于人物外貌吸引力的争论，这不仅与鲁尼作为女性小说家的性别立场有关，也与这些故事的审美愉悦根基有关：美感往往驱动着叙事。鲁尼的前两部小说也被改编成由BBC联合制作的两部电视剧，并被改编成流媒体内容。2020年的电视剧版《普通人》或许可以被解读为一系列软色情GIF动图，如果Tumblr平台在2018年没有禁止成人内容的话，这些动图在Tumblr上会非常流行。鲁尼本人和许多千禧一代文人一样，也在推特上活跃，并且习惯了各种讨论。在成为公众人物之后，她离开了这个平台；她没有想要这么多粉丝。然而，她的小说与网络的平行存在密不可分。

All three novels inspired debate over how physically attractive the characters were or should be, a fact that has to do with Rooney’s gendered position as a female novelist but also the stories’ fundamental basis in aesthetic pleasure: beauty often drives the narrative. Rooney’s first two novels have also been transformed into literal streaming content in two television series coproduced by the BBC. The 2020 TV iteration of Normal People might be best interpreted as a series of softcore-pornographic GIF sets that would have been extremely popular on Tumblr had the platform not banned adult material in 2018. Rooney was on Twitter herself, as many millennial literati were, and was accustomed to the flow of discourse. She left the platform after she became too much of a public figure; she didn’t want so many followers. Still, her novels are inextricable from their parallel existence online.

Filterworld 的文化生态系统本末倒置：推广和营销的需求凌驾于推广本身之上。文化不仅需要被设计成能够生成外部内容，以便在数字平台上进行营销；平台也能从新内容带来的参与度提升中获益。这既可以被视为一种共生关系，也可以被视为一个恶性循环，强化了迎合平台审美需求的必要性。优化这个等式——提前对创作过程进行事后评估——比寻找替代方案要容易得多。在这一点上，许多当代文化作品变得与社交平台本身相似或美化——以便更好地通过它们进行传播。

The cultural ecosystem of Filterworld puts the cart before the horse: The needs of promotion and marketing supersede the object that is meant to be promoted. Not only does culture have to be designed to generate external content to serve as marketing on digital platforms; the platforms also profit from the increased engagement driven by new content. It can be seen as either a symbiotic relationship or a vicious cycle, reinforcing the need to cater to the aesthetic requirements of the platform. Optimizing for this equation—second-guessing a creative process in advance—is much easier than finding an alternative to it. At this point, many pieces of contemporary culture have come to resemble or glorify the social platforms themselves—all the better to be distributed through them.

影响者的崛起

THE RISE OF INFLUENCERS

一部流媒体电视剧集中体现了算法平台时代文化的扁平化。它于2020年10月在Netflix上首播，当时正值疫情期间，全球大多数人仍然困在家中，除了看电视别无选择。然而，这部剧迅速占据了人们的讨论话题（至少是网络上，所有非家庭话题都发生在那里）却令人惊讶。《艾米丽在巴黎》由达伦·斯塔尔执导，他以《欲望都市》系列而闻名，该系列作品时而光鲜亮丽，时而又粗犷地刻画了20世纪90年代末的曼哈顿生活。《艾米丽在巴黎》最初计划在有线电视播出，但最终登陆了Netflix，人们可以立即观看整季，尽情狂欢。

One streaming television show epitomized the flattening of culture in the algorithmic platform era. It debuted on Netflix in October of 2020, during a pandemic that still had most people the world over stuck in their homes with no other choice but to watch TV. Still, it was surprising how quickly the show came to dominate conversation (online, at least, where all nondomestic discussion was happening). Emily in Paris was the creation of Darren Star, the director and producer best known for the Sex and the City series, an intermittently glossy and gritty portrayal of Manhattan life in the late 1990s. Emily in Paris was originally meant for cable television but eventually landed at the streaming service, where the entire first season was available immediately to binge-watch.

这部剧既是一部关于社交媒体的剧，也是一部关于社交媒体的剧。剧中主角艾米丽·库珀（由莉莉·柯林斯饰演，甜腻至极）是一位来自芝加哥的二十出头的女性。她在一家营销公司工作，扎着长长的马尾辫，喜欢穿着运动休闲服在城市里慢跑，这些都让她被称为“普通人”。艾米丽前往巴黎，将她在美国的经验带到Savoir——一家专门营销奢侈时尚品牌的虚构机构。具体来说，艾米丽的工作是培训员工如何为互联网创作内容。就像一个作为一名善意的传教士，她打破了法国传统的相对平静——长时间的午餐、平面杂志广告、时装秀——并试图让当地人转而关注 Instagram 帖子。

A show both of and about social media, the titular Emily Cooper—played saccharine in the extreme by Lily Collins—is a woman in her early twenties from Chicago. She might be called “basic” for her job at a marketing firm, her long ponytail, and her habit of jogging around cities in athleisure. Emily heads to Paris to bring her American expertise to Savoir, a fictional agency that specializes in marketing luxury fashion brands. Specifically, it is Emily’s job to train the staff on creating content for the Internet. Like a well-meaning missionary, she disrupts the relative peace of French tradition—long lunches, print magazine ads, runway shows—and tries to convert the locals to Instagram posts instead.

在星报的《欲望都市》中，凯莉·布拉德肖是一位报纸专栏作家，她撰写关于自己的约会经历、购物习惯和深厚友谊的文章——这些文章只是在她坐在笔记本电脑前时，一些伤感的草稿中略显牵强。布拉德肖的作家身份使她成为文化的重要组成部分：她构建了一种独特的个人生活和爱情哲学。相比之下，艾米丽只是一个专业的消费者。她对凯莉写作的理解就是用手机拍照，而手机在剧中也经常出现。

In Star’s Sex and the City, Carrie Bradshaw was a newspaper columnist who wrote about her dating exploits, shopping habits, and deep friendships—articles that were only hinted at in maudlin draft lines that appeared as she sat at her laptop. Bradshaw’s role as a writer made her a productive part of culture: she was constructing a particular personal philosophy of life and love. Emily, by contrast, is simply a professional consumer. Her version of Carrie’s writing is taking photos with her phone, which is a constant presence in the show.

在第一集中，她拿出手机，从自己的闺房自拍了一张照片。然后，她手机屏幕的符号图像出现在相机的真实取景框内。艾米丽将照片发布到一个未具名的Instagram账号上（以避免侵犯版权），她在Instagram上的粉丝数量也出现在了屏幕上。她将自己的用户名从@emilycooper改为@emilyinparis。在同一集的后面，她又拍了一张巴黎自拍照——她的粉丝数量翻了两番，而当她发布这张照片时，粉丝数量再次飙升。她对这个数字感到既困惑又满意。在第二集中，她发布了一些市场快照，粉丝数量增加了十倍。她照片上的点赞和评论数量像老虎机上的转轮一样快速增长：她中了社交媒体的头奖，陌生人都成了她私生活的亲密粉丝。

In the first episode, she pulls it out and takes a selfie with the view from her chambre de bonne. Then a symbolic representation of her phone screen pops up within the camera’s real-life frame. Emily posts the photo to an unnamed Instagram equivalent (to avoid copyright infringement), and her number of followers on the generic platform appears on screen. She changes her username from @emilycooper to @emilyinparis. Later in the same episode, she takes another Parisian selfie—her follower count has quadrupled, and then it jumps up again when she posts the image. She reacts to the number with bemused satisfaction. In the second episode, her followers have gone up by a factor of ten as she posts snapshots of a market. The number of likes and comments on her images spin upward like the wheels on a slot machine: she has hit the social media jackpot, acquiring strangers as intimate fans of her personal life.

这种视觉噱头贯穿了整部剧，艾米丽不停地自拍、记录派对、炫耀着自己的穿搭。剧中既没有创造任何新意，也没有传达任何深刻的见解；随着艾米丽在成长过程中不断积累粉丝，帖子也随之不断累积。与艺术摄影师在按下快门前费力构图不同，艾米丽在Instagram上拍照似乎只是瞬间完成，毫无顾忌。她只是凭着本能行事，这种本能早已被社交媒体的消费所影响，她将自己的生活实时转化为优化的内容。

This visual gimmick persists throughout the show as Emily takes yet more selfies, documents parties, and shows off outfits. Nothing new is created and no insight is imparted; the posts just rack up while the character progresses through her bildungsroman of accruing followers. Unlike, say, an art photographer laboriously framing a composition before pressing the shutter, the act of taking a photo for Instagram appears to be instantaneous and thoughtless for Emily. She just acts on instinct, already conditioned by her consumption of social media, transforming her life into optimized content in real time.

初看这部剧时，它对社交媒体的视觉化和价值化方式，让我感觉既反乌托邦又反现实。我们网络存在的幽灵时刻萦绕在心头，迫使我们不断评判生活中哪些场景值得向自己的受众群体传播，并将它们推送到动态消息中。更糟糕的是，《艾米丽在巴黎》美化了生活向内容的转化，将其塑造成主人公性格发展的证据。艾米丽在Savoir的成功源于她对社交媒体的娴熟运用；她为时尚品牌创作内容的专业工作，与她通过同一媒介构建的自我意识之间，并无任何界限。她是一个独一无二的商业品牌，并因此而感到快乐。

When I first watched the series, the way it visualized and valorized social media struck me as dystopian even as it was completely realistic. The specter of our online presences haunts every moment, causing us to constantly judge which scenes in our lives are worth broadcasting to our personal audiences, offering them up to the feed. Worse still, Emily in Paris glamorized the transformation of life into content, casting it as evidence of the protagonist’s character development. Emily succeeds at Savoir because she is so expert at social media; there is no line between her professional labor, creating content for fashion brands, and her own sense of self, which is constructed through the same medium. She is a commercial brand of one, and happy because of it.

这部剧表面光鲜亮丽，背后却隐藏着一系列预设。艾米丽与生俱来的特权却被刻意忽略，仿佛这早已是家常便饭。她是白人，身材异常纤瘦，拥有传统意义上的魅力，而且总是化着浓妆。脱离背景来看，她就像建筑师雷姆·库哈斯笔下的怪诞“正常人的幻觉”。艾米丽每天都穿着精致昂贵的服装——与其说是为了穿去逛街，不如说是为了拍照——剧中丝毫没有暗示这些奢侈服装的来源和购买资金。（艾米丽仿佛是个异想天开的贵族，想做什么就做什么，结果却把这些钱都发到了网上。）剧中其他成员也大多是白人，从艾米丽的初恋到她的老板，再到大多数同事。这实在是对一座国际化都市的拙劣描绘。不过，真实的巴黎大多只是艾米丽自拍的背景。

Beneath the show’s smoothed-out surface are a range of buried assumptions. Emily’s inherent privilege goes unremarked upon, as if it were simply a fact of life. She is white, extraordinarily thin, conventionally attractive, and always wearing heavy makeup. Considered out of context, she is a bizarre “hallucination of the normal,” as the architect Rem Koolhaas wrote. Emily wears elaborate, expensive outfits—made more for photographing than walking around in—every day, without the show giving any hint of where the luxury clothing or the money to buy them came from. (It’s as if Emily is a whimsical aristocrat, free to do whatever she pleases, which turns out to be posting online.) The rest of the cast is very white, too, from Emily’s primary love interest to her boss and most of her colleagues. It’s a poor depiction of an international city. But then, real Paris mostly functions as a backdrop for Emily’s selfies.

在第一季中，艾米丽成为了一名网红——这个词指的是那些并非凭借任何伟大的创意成就而出名的人，而是凭借其庞大的粉丝群和在网上点赞的能力而闻名。最终，这个角色获得了Savoir客户的推广合同，这一胜利被奉为值得庆祝的典范，就像乐队签约唱片公司或作家获得图书合约一样。市场营销取代了创造力，就像迎合算法信息流取代了创造性的自我表达一样——两者都不特别注重原创性。《巴黎的艾米丽》意识到生活已被社交媒体信息流所占据，以及对任何文化——时尚——的需求。生产线、零售店、公共艺术装置——以此激发内容创作。该剧记录了我们扁平化的文化，但并未对其进行批判；相反，它加速了这种扁平化。到了第三季，每个主要情节似乎都在Instagram帖子中得到解决，而且有几集还出现了来自现实生活中品牌的明显植入式广告，包括麦当劳、迈凯伦和AMI Paris：该剧的虚构营销变成了字面营销。

During the first season, Emily becomes an influencer—the term for a person known not for any great creative feat but for the number of their followers and their ability to inspire likes online. The character eventually gets included in deals to promote Savoir’s clients, a victory presented as worth celebrating the way that a band signing to a record label or an author getting a book deal once was. Marketing has taken the place of creation in the same way that conforming to the algorithmic feed has taken the place of creative self-expression—neither is particularly concerned with originality. Emily in Paris is cognizant of how life has been taken over by social media feeds and the need for any piece of culture—fashion line, retail store, public art installation—to spark content creation. The show documents our flattened culture without critiquing it; rather, it accelerates the flattening. By season three, every major plot point seemed to resolve in an Instagram post, and several episodes had blatant product placement from real-life brands, including McDonald’s, McLaren, and AMI Paris: the show’s fictional marketing became literal marketing.

节目颂扬网红身份自有其道理；2022年，一项美国调查发现，54%的受访者（年龄在13岁至38岁之间）表示，如果有机会，他们愿意成为网红。2019年另一项针对3000名儿童的调查发现，30%的儿童会选择成为YouTuber（另一种网红），而不是职业运动员、音乐家和宇航员等其他职业。在2010年代，网红这一形象被创造出来并不断传播，或许已成为文化版图上唯一主导的主角。毕竟，如果如今文化的主体发生在社交媒体上，还有什么比掌控注意力流向并以特定方式引导受众更强大的角色呢？（网红是算法的牛仔。）

That the show celebrated being an influencer made a kind of sense; in 2022, a United States survey found that 54 percent of the respondents, ages thirteen to thirty-eight, would become influencers if given the opportunity. Another survey of three thousand children in 2019 found that 30 percent of them would choose to be a YouTuber—another kind of influencer—over other careers like professional athlete, musician, and astronaut. In the 2010s, the figure of the influencer was both invented and metastasized, becoming perhaps the single dominant protagonist of the cultural landscape. After all, if the bulk of culture now takes place on social media, what more powerful role could there be than controlling the flow of attention and being able to direct an audience in particular ways? (Influencers are the cowboys of algorithms.)

这个词本身的肤浅性就说明了一切：“影响力”从来都不是终点，而只是传达特定信息的一种手段。定义网红最简单的方式就是看他们如何赚钱。就像制作杂志或播客的媒体公司一样，他们通过向既有的受众投放广告来赚钱。但最初吸引受众的往往是网红的个人生活、他们美轮美奂的周边环境（以及他们自己）以及娱乐活动。他们的生活素材——有的自然而然，有的经过精心设计——被大量地记录在 Instagram 等社交媒体平台上，他们也会在这些平台上发布品牌赞助帖子。另一个区别是，与街头派发的免费报纸或发出信号的电台不同，网红并不拥有其媒介的基础设施。他们借助数字平台的能力分发内容，既可以通过广泛安装的智能手机应用程序，也可以通过应用内信息流。

The superficiality of the word itself is indicative: “influence” is never the end point, only a means of communicating a particular message. An influencer is easiest to define by how they make money. Like a media company producing magazines or podcasts, they sell advertising shown to the audiences that they have gathered. But the content that draws the audiences in in the first place is most often the influencer’s personal life, their aesthetically appealing surroundings (as well as aesthetically appealing selves) and entertaining activities. The material of their lives—in varying degrees of organic and staged—is copiously documented on social media platforms like Instagram, where they also publish sponsored posts for brands. Another difference is that unlike a free newspaper distributed on the street or a radio station sending out a signal, influencers don’t own the infrastructure of their medium. They piggyback on the digital platforms’ ability to distribute content, both through widely installed smartphone apps and through in-app feeds.

对某人的迷恋，尤其是他们的外表或个人生活，为自我推销铺平道路，这种行为早在互联网时代之前就已存在。历史上，它最常出现在皇室宫廷或城市知识分子阶层的圈子里，地理位置相近的成员之间互相观察，就像我们今天在Instagram上发布照片一样频繁。妮侬·德·朗克洛斯于1620年出生于巴黎一个富裕家庭；父亲流亡，母亲去世后，她不得不独自谋生，并发誓终身不嫁。她成为了一名交际花，拥有多位贵族情人，同时也是一位作家和反宗教哲学家。在一个特别戏剧性的事件中，她与一位长期在乡下的情人分手，搬回了巴黎。她的前夫，维拉索侯爵，很快就搬到了街对面的一所房子里监视她，提防着新来者。为了安抚他，她剪掉了一缕头发送给他作为离别礼物。这一招奏效了，但也引发了一场模仿波波头的狂热，正如贝齐·普里奥洛（Betsy Prioleau）在她的书《诱惑者》（Seductress）中所述，这种发型被称为“尼农式波波头”（cheveux à la Ninon）。这是一个出自一位顶级网红之手的美发模因。

Fascination with a person, particularly their appearance or personal life, smoothing the way to self-promotion began long before the Internet era. Historically it shows up most often in the confines of a royal court or an urban intelligentsia, where members in geographical proximity observed each other with as much daily scrutiny as we post on Instagram today. Ninon de L’Enclos was born into a wealthy family in Paris in 1620; after her father was exiled and her mother died, she was left to make a life on her own, and she committed to never marrying. She became a courtesan with a succession of noble lovers, as well as an author and anti-religious philosopher. In one particularly dramatic episode, she broke up with a long-term lover in the countryside and moved back to Paris. Her ex, the Marquis de Villarceaux, promptly moved into a house across the street to spy on her, watching for new visitors. To mollify him, she chopped off a length of her hair to send to him as a parting gift. It worked, but also launched a mania for copycat bob haircuts, called “cheveux à la Ninon,” as Betsy Prioleau recounts in her book Seductress. It was a hairdressing meme from a consummate influencer.

后来，在1882年，英国社交名媛兼演员莉莉·兰特里成为第一位代言产品的名人，她成为梨牌香皂的代言人，她的肖像被印在广告上。兰特里以其白皙的肤色而闻名，这很可能是遗传而非香皂的功效，但她的个人形象赋予了广告以权威性。事实上，兰特里正是通过复制她的形象而出名的，最初是英国著名画家的作品，后来是印在明信片上的素描。在复制品如此稀缺的时代，复制品本身就足以代表认可——图像的传播相当于点赞。

Later, in 1882, the British socialite and actor Lillie Langtry became the first celebrity to endorse a product when she became the face of Pears Soaps, her portrait reproduced on advertising. Langtry was known for her bright white complexion, likely more hereditary than an effect of the soap, but her personal image lent authority to the ads. In fact, Langtry had become famous through the reproduction of her image, first by prominent British painters and then in sketches printed on postcards. When reproduction was so scarce, its very fact could stand in for approval—the proliferation of an image the equivalent of a like.

消费者一直以来都关心那些因其他事情而出名的名人的生活方式：演员的饮食习惯、大亨的赛马、画家的风流韵事。名声——也可以定义为关注——本身就散发着光环，几乎让任何事物都变得有趣。20世纪60年代，安迪·沃霍尔用他的“超级明星”系列作品将名声的平庸推向了极致。在20世纪60年代初的沃霍尔“试镜”系列中，一群在圈外并不知名的反主流文化人物在摄像机前摆出长时间的姿势，镜头只是观察着他们的脸，就像在为一部不存在的电影试镜一样。（如今，他们带来了（例如，在自拍前进行心理排练，或在为 TikTok 录制正面视频之前的瞬间。）这些视频表明，我们只需要凝视某人就能感觉到他们很重要——相机镜头和投影仪的取景会产生魔力。

Consumers have always cared about the lifestyle decisions of celebrities famous for something else: the actor’s diet regimen, the tycoon’s racehorses, the painter’s affairs. Fame—which could also be defined as attention—casts its own aura, making just about anything interesting. In the 1960s, Andy Warhol took the banality of fame to its end point with his group of “Superstars.” For Warhol’s “Screen Test” series in the early 1960s, a set of countercultural characters, not well known outside of their milieu, posed in front of a video camera for extended durations, the lens simply observing their faces, like an audition for a nonexistent film. (Today, they bring to mind rehearsals for selfies or the moments before a front-facing video is recorded for TikTok.) The videos demonstrate that we need do nothing more than gaze at someone to feel as though they are important—the framing of the camera lens and the projector casts a spell.

电脑或手机屏幕也能实现同样的效果：TikTok动态中的任何人都会显得和其他人一样出名，这又是另一种扁平化。沃霍尔曾说过：“未来每个人都会因为15分钟而举世闻名。” 1991年，音乐家兼在线博客先驱Momus颇具先见之明地将这一预言更新为：“未来每个人都会因为15个人而出名。” 在社交媒体上，这已成为事实。每个用户都因其粉丝而出名。

The computer or phone screen accomplishes the same feat: anyone within a TikTok feed appears just as famous as everyone else, another form of flattening. Warhol once said, “In the future everyone will be world-famous for 15 minutes.” In 1991, the musician and online blogging pioneer Momus presciently updated the prediction to “In the future everyone will be famous for 15 people.” On social media, that has simply become fact. Every user is famous to their followers.

网红可以说是博主的继承者，后者是21世纪初新兴主流互联网的明星。博主们同样利用数字工具自行发布，并通过文字快讯将个人生活戏剧化——其中最典型的代表是2002年左右兴起的“妈妈博主”浪潮。博主们也通过赞助帖和网站边缘的横幅广告来赚取广告费。根据谷歌趋势追踪系统（记录搜索词的流行度）的数据，“博主”一词自2011年以来一直在缓慢下滑，“网红”一词则在2016年初强势崛起。当时，Instagram用户数突破5亿，社交网络信息流也普遍变得更加算法化。Instagram为网红提供了一个完美的平台：无需长篇大论的文字或自我情感的表露。仅凭精美的图片就足以吸引粉丝，并最终实现盈利。“博主”一词描述的是字面意义上的写作活动，而“影响力”则更接近于经济层面。它是一种销售工作，说服受众购买某种东西，首先是理想生活方式的愿景，然后是构成这种愿景的产品。

The influencer is something of a successor to the blogger, the star of the nascent mainstream Internet in the 2000s. Bloggers similarly used digital tools to self-publish and dramatize their personal lives in running textual dispatches—the wave of “mommy bloggers” starting around 2002 foremost among them. Bloggers sold advertising, too, in the form of sponsored posts and banner ads that appeared on the borders of their websites. According to Google’s trend tracker, which documents the popularity of search terms, “blogger” has been on a slow slide downward since 2011, while “influencer” emerged in strength around the beginning of 2016. That was also when Instagram passed five hundred million users and social-network feeds were generally becoming more algorithmic. Instagram offered a perfect venue for the influencer: no voluminous writing or emotional disclosure of the self were required. Glossy images alone were enough to attract a following that could later be monetized. Where the term blogger described a literal activity of writing, “influencing” is closer to the financial side of what’s going on. It’s a sales job, convincing audiences to buy something, first a vision of aspirational lifestyle and then the products that make it up.

Patrick Janelle 可以说是 Netflix 的 Emily Cooper 的现实版，尽管他的网红生涯比她早得多。在 2010 年代初期到中期，他将自己的生活转向了内容创作，并逐渐以此为基础建立了自己的事业，之后又创立了自己的公司。虽然如今，他的Instagram账号@aguynamedpatrick已不再是他的主要关注点，但如今仍然拥有超过四十万粉丝和六千多张照片，这些照片都是他十年来一直使用的。他的账号@aguynamedpatrick平易近人。他发布的照片​​就像《GQ》之类的男性杂志的写真集：富丽堂皇的酒店房间、通风良好的咖啡馆、曼哈顿公寓里装饰精美的艺术墙，所有这些都以粗犷英俊的珍妮尔为主角。所有照片都清晰锐利，几何构图，都是珍妮尔亲自用iPhone拍摄的，有时还会借助遥控快门。然而，他的生活并非一直如此精心打扮，尤其是在2012年他开始使用Instagram之后。

Patrick Janelle is something of a real-life version of Netflix’s Emily Cooper, though his career as an influencer started long before hers. In the early to mid-2010s, he turned his life into content and gradually built a career, and then businesses, from it. Though it’s no longer his primary focus, his Instagram account today still has over four hundred thousand followers and over six thousand images from a decade of use under the approachable account name @aguynamedpatrick. His posts resemble photo shoots from a men’s magazine like GQ: palatial hotel rooms, airy coffee shops, the art-bedecked walls of his well-furnished Manhattan apartment, all starring the ruggedly handsome Janelle as the main character. Crisp, clear, and geometrically composed, all the photos are taken by Janelle himself with an iPhone, sometimes with the help of a remote trigger. His life wasn’t always so manicured, however, particularly when he got on Instagram in 2012.

我第一次见到珍妮尔是在几年前，在苏荷区他公寓附近一个街角，他推荐了一家叫“The Dutch”的餐厅。餐厅里配备了必备的球形吊灯和开放式搁架。这家餐厅可以说是珍妮尔的总部，他甚至把它放在了自己的网站上。2011年，他刚满30岁，搬到纽​​约，在美食杂志《Bon Appétit》做自由设计师。Instagram成了他与朋友们分享生活的平台，这些朋友们来自他出国旅行的家乡，也来自他长大的科罗拉多州。“这个媒介就是用来记录生活的，”他告诉我。他拥有职业模特般沉稳的自信。在去杂志社工作之前，珍妮尔经常光顾曼哈顿的咖啡馆，并在Instagram上创建了一个标签#dailycortado，指的是当时咖啡爱好者最爱的小杯卡布奇诺。他每天都会拍一张他那杯抹着拿铁咖啡的照片，咖啡放在大理石或木质桌子上——这些手工打造的桌面恰好与光滑的瓷质咖啡杯形成了鲜明的对比——然后发到Instagram上。这些材质象征着奢华与品质；它们的自然质感与智能手机屏幕的光滑平整形成了鲜明的对比，尽管这些照片本身也成为了数码素材。

I first met Janelle years ago at The Dutch, a restaurant he suggested on a corner in SoHo near his apartment. It was outfitted with the requisite orb pendant lamps and open shelving. The restaurant was something of a headquarters for Janelle, who even featured it on his website. In 2011, he had just turned thirty, moved to New York, and taken a freelance gig as a designer at the food magazine Bon Appétit. Instagram became a way to share his life with friends, both from his travels abroad and back where he grew up in Colorado. “The medium was built for documenting your life,” he told me. He had the placid confidence of a professional model. Frequenting Manhattan cafés before magazine work, Janelle started an Instagram hashtag, #dailycortado, referring to the short cappuccino that was a favorite of coffee aficionados at the time. Every day, he would take a photo of his latte-art-topped coffee resting on a marble or wood table—artisanal surfaces that also happened to contrast nicely with a smooth porcelain coffee cup—and post it on Instagram. The materials were signs of expense and quality; their naturalness and texture created a compelling contrast with the slick flatness of the smartphone screen, though the photographs were so much digital fodder, too.

Janelle 是 Instagram 上这种审美的先驱之一，她帮助将这种审美牢牢地植入到 Instagram 上，使其成为高雅的普通咖啡馆设计的标志。他还推广了这个标签，让其他人也使用，像任何好的表情包一样，邀请大家参与其中。十年过去了，#dailycortado 依然活跃，成为 Instagram 平台的基石。现在，它已经融入了 Instagram。在用户堆成标题的标签列表中，以便更多人可以发现他们的帐户，从而促使算法向咖啡爱好者推荐照片。

Janelle was one of the pioneers of this aesthetic on Instagram, helping to firmly embed it as the signal of tasteful generic café design. He also popularized that hashtag for other people to use, inviting participation as any good meme does. A decade later, #dailycortado is still active, part of the bedrock of the platform. Now, it’s included in lists of hashtags that users pile into captions so that more people can discover their accounts, goosing the algorithm to recommend the photo to coffee fans.

如果我们回顾这些照片，会发现它们有一种矛盾的特质。珍妮尔捕捉的是放慢脚步、专注的瞬间：放下咖啡杯，坐下来，在畅饮前欣赏片刻。这是喧嚣都市生活中的片刻宁静；街上其他人匆匆而过，而你却静静地待着。然而，每一条 Instagram 帖子都为信息流这头猛兽提供能量，信息流以最快的速度将尽可能多的内容传送到你的眼前——这与放慢速度截然相反。相反，照片是加速的载体，既是为了内容消费，也是为了扩大珍妮尔的网络影响力，扩大他的受众群体，让他作为网红赚钱。珍妮尔可以放慢速度，但只是为了拍出构图完美的照片，而他的观众或许也可以在信息流将他们的注意力吸引到更多刺激之前，间接地放慢速度。

If we look back at the photos, they have a paradoxical quality. Janelle was capturing a moment of slowing down and paying attention: when you set down your coffee cup, sit down, and appreciate it for a split second before drinking. It’s a restful moment in bustling city life; you stay still while everyone else on the street careens by. And yet each Instagram post feeds the beast of the feed, which sends as much content as possible by your eyes as fast as possible—the opposite of slowing down. Instead, the photo is a vector for acceleration, both for content consumption and the expansion of Janelle’s online footprint, the audience that makes him money as an influencer. Janelle can slow down, but only to take the perfectly composed photo, and his viewers can perhaps vicariously slow down for a moment before the feed sweeps their attention on to yet more stimulation.

“一开始只是喝咖啡，看一些赏心悦目的照片；那是一种身处城市的感觉，”珍妮尔谈到他早期使用 Instagram 时说道。他发布的内容取材于街头摄影的词汇：消防梯、建筑立面、熙熙攘攘的人群。但随着时间的推移，随着 Instagram 和网红的蓬勃发展，他那些充满抱负的城市生活照片也成为了他实现这种生活方式的一种方式。奢华的照片让他置身于奢华的境地。“生活方式催生生活方式，”他告诉我。“随着我不断发布帖子，我作为参与者的需求越来越大，获得了更多机会，并在经济上建立了自己的品牌。”他通过自拍展现出的个人时尚感，为他赢得了与拉夫·劳伦的推广合约。他发布的一次度假经历让他获得了免费酒店房间的优惠。

“In the beginning it was just coffee, and images I would see that were just visually pleasing; it was about being in the city,” Janelle said of his early Instagram use. His posts were drawn from the vocabulary of street photography: fire escapes, building facades, milling crowds. But over time, and with the booming popularity of Instagram and influencers, his aspirational photos of city living became a way to achieve that same lifestyle. Images of luxury led him into luxury situations. “Lifestyle begot lifestyle,” he told me. “As I continued to post, I was able to become more in demand as a participant in things, having more access to opportunities, building out a brand financially.” His personal fashion sense, shown off in selfies, led to a promotional deal with Ralph Lauren. A vacation that he posted led to offers of free hotel rooms.

影响力成了我的职业，尽管珍妮尔不喜欢“潮流引领者”这样的词。“我不认为我在创造它，创造潮流，而是在选择我认为新鲜刺激的东西，并记录下来，”他说。2014年，美国时装设计师协会将珍妮尔评选为首位“年度Instagram用户”。2015年的一天，我在Instagram探索页面上发现了珍妮尔，这个页面收集了算法推荐的帖子，推荐点赞或关注账户。值得关注。我猜，我关注了足够多的纽约餐厅和室内设计灵感账号，里面全是市中心的阁楼和中世纪家具，所以 Janelle 的账号完全符合我的品味，至少根据算法是这样。

Influencing became a career, though Janelle dislikes words like “tastemaker.” “I don’t think I’m creating it, making the taste, but I am selecting the things I think are new and exciting, documenting them,” he said. In 2014, the Council of Fashion Designers of America named Janelle its first “Instagrammer of the Year.” I found Janelle one day in 2015 on my Instagram Explore page, which collected algorithmic recommendations of posts to like or accounts to follow. My guess is that I was following enough New York City restaurants and interior design inspiration accounts, filled with downtown lofts and mid-century furniture, that Janelle’s account slotted squarely into my perceived taste, at least according to the algorithm.

我当时快三十岁了，住在布鲁克林，正在寻找成年人生活应该是什么样子——本质上，就是如何以及在哪里花掉我那点儿多余的钱。珍妮尔精心构图和挑选的照片多少透露出一些这种感觉，尽管他逐渐演变成一个沉迷于奢侈品的浪荡公子，不如他的街景快照那么容易让人产生共鸣。Instagram 也为这些生活方式内容披上了一层真实的外衣，而媒体公司制作的纸质杂志对我来说是无法拥有这种真实感的——尽管他的帖子在某种程度上是企业化的，因为它们是通过 Instagram 这家科技公司过滤的。“我知道我是一个有抱负的人，”珍妮尔告诉我。“通过发布照片并让观众看到，我实际上是在暗示，‘我支持这个。’我想激励人们为自己创造尽可能美好的生活。”

I was also entering my late twenties as a Brooklyn resident and looking for some definition of what adult life should look like—in essence, how and where to spend the little extra money I had. Janelle’s composed and curated photos gave a hint of that, though his evolving persona as a luxury-focused gadabout was less relatable than his street-scene snapshots. Instagram also lent a veneer of authenticity to such lifestyle content that a print magazine produced by a media corporation wouldn’t have had for me—though his posts were corporate in a different way, since they were filtered through Instagram as a tech company. “I understand that I’m an aspirational figure,” Janelle told me. “By me putting up an image and giving it some exposure to an audience, I’m implicitly saying, ‘I endorse this.’ I want to sort of jump-start people into creating for themselves the best life possible.”

阶级问题贯穿了珍妮尔的整个人生轨迹：他模仿上流社会的消费主义形象，并以此达成目标，强化了他们的地位象征，就像许多网红一样。（影响力源于嫉妒。）珍妮尔身材苗条，肤色白皙，传统意义上的魅力，使她能够轻松地融入他为自己打造的奢华形象。某种程度上，他很容易被算法曝光，因为他的外貌如此受欢迎，符合城市富人的既定视觉词汇。就在珍妮尔不断积累粉丝的同时，其他人的账号却被推荐算法限制或被严厉的审查制度屏蔽：发布政治煽动表情包的活动人士、自我宣传的性工作者，以及那些想像珍妮尔一样在Instagram上裸露上身的女性（这种差异激发了“解放乳头”运动的兴起）。

There’s a persistent issue of class throughout Janelle’s path: He enacts upper-class consumerist tropes and in doing so achieves them, reinforcing their status signaling, as many influencers do. (Influencing runs on jealousy.) White, traditionally attractive, and thin, Janelle was able to easily inhabit the luxurious imagery he made for himself. In a way, he was primed for algorithmic exposure because his appearance was so acceptable, fitting within the established visual vocabulary of urban wealth. Even as Janelle was gaining followers, others found their accounts throttled by the recommendation algorithm or blocked by aggressive censorship: activists posting political agitation memes, sex workers promoting themselves, and women who wanted to be as topless on Instagram as Janelle often was (a discrepancy that inspired the Free the Nipple campaign).

多年来，随着社交媒体的蓬勃发展，珍妮尔几乎成了历史人物。他的Instagram粉丝在2010年代达到顶峰，而那时他的审美似乎已经停滞：那是一个充斥着名字里带“&”符号的仿工业风餐厅、费力调制的手工鸡尾酒和学院风男装的时代。时尚（每个人都看起来像参加婚礼的伐木工人）。事实上，Janelle 提到了他与 Instagram 关系的一个回顾性转折点，尽管他仍然在 Instagram 上相当活跃：2016 年，当时它的动态从按时间顺序排列变成了按算法排列。

Over the years, Janelle has become something of a historical figure as the march of social media has passed him on. His Instagram following peaked in the 2010s, which is when his aesthetic feels frozen: it was the era of faux-industrial restaurants with ampersands in the name, arduously mixed craft cocktails, and preppily neat men’s fashion (everyone looked like a lumberjack attending a wedding). In fact, Janelle cites a retrospective turning point in his relationship with Instagram, though he is still quite active there: 2016, when its feed changed from chronological to algorithmic.

“就我个人而言，在我通过社交媒体成长的过程中，平台的改变赋予算法权力是最糟糕的事情，”他告诉我。起初，他可以每天发布几条帖子，并且知道这些帖子会实时地、按照符合他叙事逻辑的时间顺序推送到粉丝手中。现在，他无法预知自己的图片何时、以何种顺序出现。这不仅对消费方，也对创作方都令人沮丧：“没有人真正拥有自主选择的权利，没有人能够以自己喜欢的方式去策划他们所看到的内容。一切都被算法所左右。”

“For me, personally, in the way that I grew through social media, the change of the platform to give power to the algorithm was worst possible thing,” he told me. Originally, he could post several times a day and know that his posts would reach his followers in real time, in a chronological order that made sense for his storytelling. Now he can’t tell when or in what order his images will show up. It’s frustrating on the consumption side as well as creation: “No one is really given that choice to make that decision for themselves, to curate what they’re seeing in a way that feels great to them. It’s all being dictated by the algorithm.”

Instagram 多年来也不断改变着自己的风格，从静态图片信息流转向了短视频，模仿 Snapchat；Instagram TV 则专注于长视频；最终，Reels 则模仿 TikTok，专注于短视频。这款应用逐渐失去了其作为表达个人品味的相对朴素空间的定位。“这些决定绝不是为了让平台成为内容创作者更好的平台，”Janelle 说，“它们纯粹是增长团队和业务团队为了拓展业务而做出的决定。”

Instagram shifted formats over the years, too, moving away from a feed of still images and adding stories, for ephemeral posts that mimicked Snapchat; Instagram TV, for longer videos; and eventually Reels, for short videos to copy TikTok. The app gradually lost its identity as a relatively austere space for expressing your own taste. “At no point were these decisions made to make the platform a better place for the people creating the content,” Janelle said. “They were purely decisions made by the growth team and by the business team to figure out how they could expand.”

Instagram 设计不断变化的特性意味着，几乎不可能在 Janelle 作品最初存在的语境中体验它。你无法倒回时间，看看这个账号在 2015 年是如何呈现的，比如在网格或时间线上。数字平台缺乏对文化长寿至关重要的那种稳定性，就像你可能会看到摄影师几十年前的胶卷底片在博物馆里展出一样。随着公司优先级的改变，语境也被一扫而空。“不仅事物在当下发生变化，而且由于所有这些都是数字化的，它可能会追溯性地改变，因为我们并不拥有这些东西。我们根本无法控制它如何呈现或使用，”Janelle 说。Instagram 的动态从单纯的方形图片变成了任意大小的图片，然后开始包含与静态照片冲突的视频，所有这些都是为了竞相提供更引人注目、更快、更个性化的内容。

The ever-shifting nature of Instagram’s design means it is almost impossible to experience Janelle’s work in the context in which it originally existed. You can’t rewind time and see how the account was presented in 2015, for example, on the grid or in the timeline. Digital platforms lack the kind of stability that is vital for the longevity of culture, the way you might see a photographer’s decades-old film negatives displayed in a museum. Context is swept away as the companies change priorities. “Not only are things changing in the moment, but because all of this is digital, it could be changed retroactively, because we don’t own the thing. We have no control over how it’s presented or used at all,” Janelle said. The Instagram feed changed from solely square images to any size, and then began including videos that clashed with still photographs, all in the name of competing to deliver more compelling content, faster and more personalized.

Janelle 对 Instagram 感到不满，但也不想切换到 TikTok 这样的新平台——在如今的网红经济环境下，想要生存下去，就必须玩跳房子游戏。他知道 TikTok 或许能提供更好的商业机会，但也会耗费他太多的创作精力。（不过，更新动态的吸引力或许难以抗拒：我们交谈几个月后，他开始尝试性地发布 TikTok 视频，并在那里积累粉丝。）这种不断探索下一个热门社交媒体平台的需求，让人想起 20 世纪初默片明星试图转战有声电影，或者戏剧演员转战电视：并非每个人都能成功，也并非每个人的艺术创作方式都能在新媒体中发挥作用。

Janelle became disaffected with Instagram but also didn’t want to switch to newer alternatives like TikTok—a game of hopscotch that is often required to survive in the current influencer economy. He knew TikTok might offer better business opportunities, but it also would have taken up too much of his creative energy. (Still, the gravity of the latest feed might be irresistible: months after we spoke, he began tentatively posting TikTok videos and accruing followers there, too.) The constant need to figure out the next big social media platform is reminiscent of early silent-film stars trying to make the switch to talkies in the early twentieth century, or theater actors moving to television: not everyone made it, and not everyone’s artistic approach functioned in the new medium.

如今，转型需要每隔几年（甚至每年）而不是几十年一次，这需要对科技垃圾有更强的敏感度。你必须知道何时避免彻底失败，比如昙花一现的移动视频流媒体服务 Quibi（“quick bites”的缩写）。尽管内容预算高昂，但用户却寥寥无几，不到一年就关门大吉了。很多炒作最终都以失败告终，网红在一个失败的平台上投入的所有努力都付诸东流。

The need to pivot now happens every few years (if not annually) instead of decades and requires a stronger sensitivity to technological bullshit. You must know when to avoid the total duds, like the short-lived mobile video streaming service Quibi (short for “quick bites”), which drew few users despite its high-budget content and shut down after less than a year. Much of the hype doesn’t pan out, and any effort an influencer invests in a failed platform is lost.

Janelle 是一位精明的企业家。Filterworld 的潮流瞬息万变，因此他没有不断地调整自己的内容以适应各种流行的信息流格式，而是创办了一家名为“Untitled Secret”的公司，代理十几位其他网红，并在不断壮大的员工队伍的帮助下，协调与企业客户的营销交易。他影响着这些网红。

Janelle is a savvy entrepreneur. Fads shift quickly in Filterworld, so rather than continually adapting his own content to whichever formats feeds favored, he built a company called Untitled Secret that represents a dozen other influencers and coordinates marketing deals with corporate clients, helped by a growing staff. He influences the influencers.

虽然珍妮尔所推崇的2010年代工业风依然盛行，但前卫的品味已经转向了更混乱、更无序，甚至带有强烈的不真实感。如今，网红们通常拥有数百万粉丝，在这样的规模下，建立个人联系几乎是不可能的，这与传统名人的疏离感如出一辙。卡戴珊式的真人秀节目催生了一批新的网红，他们在小屏幕和更小的手机屏幕之间游刃有余：你可以在电视上看到金，然后在Instagram上关注她，她现在在Instagram上有349个粉丝。百万粉丝，一个分散的粉丝群体。与珍妮尔的情况不同，Instagram 并非卡戴珊名气的源泉，而只是另一个承载名气的容器。已有的名气是启动算法推广的最佳途径，而算法推广又能进一步增强名气。

While the overall industrial-chic vibe of the 2010s that Janelle espoused is still very prevalent, cutting-edge taste has moved on to the messier and more chaotic, even the aggressively inauthentic. Influencers now regularly have millions of followers, a scale at which personal connection is barely possible, replicating the remove of traditional celebrity. Kardashian-style reality television ushered in a new class of influencer who moved smoothly between the small screen and the even smaller one of phones: You could watch Kim on TV and then follow her on Instagram, where she now has 349 million followers, the population of a dispersed nation of fans. Unlike Janelle’s case, Instagram wasn’t the source of Kardashian fame so much as just another container for it. Preexisting fame is the best way to kick-start algorithmic promotion, which further enhances it.

虽然社交媒体早期的愿景是将用户与真实的朋友联系起来，但随着时间的推移，不真实也逐渐成为一种可以接受的事物。2016年，科技公司Brud的联合创始人Trevor McFedries和Sara DeCou创建了一个名为@lilmiquela（小米克拉）的Instagram账户。账户简介描述了一位19岁的巴西裔美国女孩，其照片也都是Instagram的标配：随意的自拍、与朋友的合影，以及在城墙前摆拍的肖像。米克拉的审美是普通的网红，与她之前成功账户的平均水平一致。她自己也很普通：不是真人，而是电脑渲染的。

While the early promise of social media was to connect users to their actual friends, over time inauthenticity became something to embrace. In 2016, Trevor McFedries and Sara DeCou, cofounders of a technology company called Brud, created an Instagram account named @lilmiquela (Little Miquela). The account’s bio described a Brazilian American nineteen-year-old, and its images were the standard Instagram fare: casual selfies, snapshots with friends, and posed portraits in front of city walls. Miquela’s aesthetic was generic influencer, the average of successful accounts before her. She was generic, too: not a human but a computer rendering.

她的肌肤如蜡般光滑，由内而外散发着光泽。她的眼神空洞得有些奇怪。然而，她出现在真实的人身边，身着真实的服饰，身处真实的场所。这个账号的本质是一个虚拟的三维模型，一个近乎逼真的数字玩偶，可以摆出任何你能想到的姿势。这些光鲜亮丽的图像，只比大多数时尚大片略显虚假，最终取得了成功：她作为第一位著名的“虚拟网红”登上头条，并与人才经纪公司WME签约（或者至少是她的创造者签约了）。米奎拉目前在Instagram上拥有近300万粉丝，并且像任何成功的网红一样，与Calvin Klein和Prada等奢侈时尚品牌签订了推广协议。

Her skin was waxy and inhumanly smooth, glowing from within. Her eyes were strangely blank. Nevertheless, she appeared next to real people, dressed in real clothes, in real places. The essence of the account is a virtual three-dimensional model, a nigh-photorealistic digital doll that can be posed any conceivable way. The glossy imagery, only slightly faker than most fashion shoots, proved successful: she made headlines as the first famous “virtual influencer” and signed up with the talent agency WME (or, at least, her creators did). Miquela currently has nearly three million Instagram followers and, like any successful influencer, does promotional deals with luxury fashion brands like Calvin Klein and Prada.

她的存在本身就让她的影响力更具销售力，因为没有一个拥有独立人格的人来阻碍广告的传播。米奎拉不会像帕特里克·珍妮尔那样，因为信息流的变化而感到疏远，也不会无法适应下一个重要的多媒体格式。她也会永远保持19岁——直到她突然进入一个新的受众群体，这能让她更有利可图。当这个账号不再具有相关性时，它可以简单地关闭，并发展出一个新的角色。这是信息流规模越来越大、更新速度越来越快，内容可互换性带来的另一个后果：我们不一定能注意到一个内容来源被另一个内容来源替换了。曾经驱动我们社交媒体用户对帕特里克·贾内尔（Patrick Janelle）这样的人物的刻画，如今已完全被塑造成虚构的人物。如果说网红的生活早已沾染上虚构色彩——被赞助的旅行和赠送的服装装点得光鲜亮丽，而这些交易却鲜为人知，仿佛是自然而然发生的——那么网红本身也变成了虚构的人物，就像谷物品牌的卡通吉祥物一样。

That she doesn’t exist makes her influence more salable, because there is no human being with an independent personality to stand in the way of the advertising. Miquela won’t feel alienated by changes in the feed, as Patrick Janelle did, or fail to adapt to the next big multimedia format. She will also stay nineteen forever—until it becomes more profitable for her to suddenly age into a new demographic. And when the account has outlived its relevance, it can simply shut down and develop a new persona. This is another consequence of content’s interchangeability as the feeds grow larger and faster: we don’t necessarily notice when one source of content is swapped out for another. The aura of intimacy that once drove social media users to people like Patrick Janelle is now cultivated for entirely imaginary characters. If influencers’ lives were already tinged with fiction—glitzed up with sponsored travel and gifted outfits, the deals undisclosed as if they happened naturally—then the influencers have since become fictional themselves, like the cartoon mascots of cereal brands.

影响者的身份既专业化又分散。迎合信息流、积累粉丝并出售接触个人受众的渠道，是一条可行的职业道路。查莉·达梅利奥 (Charli D'Amelio) 是一位居住在康涅狄格州郊区、名不见经传的青少年舞者，她于 2019 年 5 月加入 TikTok，当时这个社交网络主要以对口型和舞蹈视频为主。通过紧跟潮流和参与编舞表情包，达梅利奥现在拥有超过 1.5 亿粉丝，这或许是 TikTok 算法推送功能有史以来增长最快的社交媒体账户。短短几年间，达梅利奥从社交媒体上的小明星一跃成为主流名人，甚至与卡戴珊家族的一位三流成员约会，这段跨界关系足以拍成一部有线电视情景喜剧。

The influencer identity has been both professionalized and dispersed. There is a viable career path in catering to the feed, accruing followers, and selling access to your personal audience. Charli D’Amelio, a relatively unremarkable teenage dancer living in suburban Connecticut, joined TikTok in May 2019, when the social network was mostly oriented around lip-synching and dancing videos. By jumping on trends and participating in choreography memes, D’Amelio now has over 150 million followers, perhaps the fastest rise of a once-obscure social media account ever, thanks to TikTok’s algorithmic feed. Over just a few years, D’Amelio went from minor social media star to fully mainstream celebrity, even dating a tertiary member of the Kardashian clan in a crossover worthy of cable sitcoms.

在这十年里，个人网红不再引人注目，也因为许多数字平台用户被迫像网红一样行事，不断创作内容，积累受众，并想方设法将其货币化——要么通过直接投放广告，要么逐渐吸引同行的关注。我们都是网红：视觉艺术家试图在Instagram上吸引画廊或策展人的注意；小说家用一连串的推文记录写作过程；业余面包师在TikTok上制作面包视频并回复评论，试图在自己的公寓里创业。还有约会网红和个人理财网红。对于Filterworld中的众多职业道路来说，遵循各种信息流的需求几乎成了不可避免的戒律。压力如此之大，以至于宣传内容有时会取代实际的创作。

Individual influencers are less remarkable in this decade also because so many users of digital platforms are pressured to act like influencers themselves, constantly creating content, accruing an audience, and figuring out ways to monetize it—either immediately through literal advertising or more gradually through the attention of their peers. We are all influencers: the visual artist trying to get the attention of galleries or curators on Instagram; the novelist documenting their writing process with a litany of tweets; the amateur baker making bread TikTok videos and replying to comments as they try to build a business out of their apartment. There are dating influencers and personal finance influencers. For so many career paths in Filterworld, following the demands of various feeds has become an almost unavoidable commandment. The pressure is so great that the promotional content has a way of superseding the actual craft.

内容资本

CONTENT CAPITAL

在 Filterworld，文化变得越来越迭代。创作者很难直接制作电影或出版书籍；她需要先发布样本素材，阐述自己的愿景，并在网上吸引那些热爱她作品的忠实粉丝。例如，一本书必须先在推特上引起热议，然后才能为引发公众讨论的优秀文章提供素材，或许还能引发后续报道或评论。读者必须转发她的想法，并在 Instagram 快拍上分享重要引言。经纪人必须注意到蓬勃发展的势头，并与这位新晋作家签约。最终，或许会有出版商考虑她的手稿——前提是她积累了足够的“平台”，即拥有足够多的社交媒体粉丝，并具备影响内容流的能力。 （这个词很能说明问题：人也是平台。）一旦书卖给出版商，并在书店上架，作者就会利用这个平台，在推特上发粉丝，上传封面图片，在抖音视频中展示这本书——在尽可能多的地方吸引读者关注它的发布。（当你拿着这本书的时候，也会有同样的感受，至少如果我做得够好的话。）

In Filterworld, culture has become increasingly iterative. It’s harder for a creator to go straight to making a movie or publishing a book; she needs to first publish her sample material, describe her vision, and gather an audience online who are engaged fans of her work. A book, for example, must first make for good tweets and then provide the material for good essays prompting public dialogue, perhaps inspiring a follow-up or an op-ed. Readers must retweet her thoughts and share pull quotes on their Instagram stories. An agent must notice the burgeoning momentum and sign the nascent author. Then finally, perhaps, a publisher will consider her manuscript—if she has accrued enough of a “platform,” that is, high enough numbers of social media followers and an ability to influence the content stream. (The choice of word is telling: people are platforms, too.) Once the book sells to a publisher and debuts on bookstore shelves, the author leverages that platform, tweeting to her fans, posting images of the cover, holding it up in a TikTok video—driving attention to its launch in as many spaces as possible. (The same will be true for this book by the time you are holding it, at least if I’m effective at it.)

这种通过在社交媒体上取得成功来提前圈定受众的需求，可以用一个很贴切的术语“内容资本”来解释。该术语由学者凯特·艾希霍恩（Kate Eichhorn）在其2022年的专著《内容》 （Content）中提出，描述了互联网时代的一种状态，即“一个人作为艺术家或作家从事工作的能力越来越取决于其内容资本；也就是说，取决于一个人创作内容的能力，这些内容并非关于他的作品，而是关于他作为艺术家、作家或表演者的身份。”换句话说，重点不在于事物本身，而在于围绕它的氛围，即一个人因过着创作者的生活方式而产生的辅助材料。这些辅助内容可能是Instagram上的自拍、绘画工作室的照片、旅行的痕迹、推特上的随笔，或是TikTok上的一段独白。所有这些都为创作者建立了受众群体，而创作者与他们所创作的事物仍然是一个独立的个体。如果罗兰·巴特 1967 年的文章预言了“作者的死亡”，那么现在作者的个人品牌才是最重要的；作品本身已经消亡。

This need to corral an audience in advance by succeeding on social media can be explained by the useful phrase “content capital.” Established by the scholar Kate Eichhorn in her 2022 monograph Content, it describes the Internet-era state in which “one’s ability to engage in work as an artist or a writer is increasingly contingent on one’s content capital; that is, on one’s ability to produce content not about one’s work but about one’s status as an artist, writer, or performer.” In other words, the emphasis is not on the thing itself but the aura that surrounds it, the ancillary material that one produces because of living the lifestyle of a creator. That ancillary content might be Instagram selfies, photos of a painting studio, evidence of travel, tossed-off observations on Twitter, or a monologue on TikTok. It all builds an audience for the person, who remains a separate entity from the things that they make. If Roland Barthes’s 1967 essay predicted “the death of the author,” the author’s personal brand is now all that matters; it’s the work itself that is dead.

艾希霍恩回应了社会学家皮埃尔·布迪厄（Pierre Bourdieu）在20世纪70年代提出的“文化资本”概念：对各种高雅文化形式的熟练掌握，这种文化能够赋予社会地位，并帮助精英阶层成员相互认同。文化资本是指知道羊绒比棉花更能激发人们的渴望，或者知道杰克逊·波洛克的画作远不止是孩子们可以临摹的一堆水滴。（布迪厄指出，在西方，对激进的美学实验和抽象主义持开放态度是精英阶层的标志。）这不仅是一种对艺术的理解，也是一种对艺术在社会语境中的意义，以及不同作品或艺术家所象征意义的理解。因此，内容资本是指对数字内容的熟练掌握：了解应该制作什么样的内容，了解各种平台的信息流如何运作，了解它们的优先级，以及受众对特定创作的反应。拥有更多内容资本的人会获得更多粉丝，从而在Filterworld的文化生态系统中拥有更大的影响力。

Eichhorn responds to the sociologist Pierre Bourdieu’s 1970s concept of “cultural capital”: the fluency in forms of high culture that could bestow social status and help members of elite classes to identify one another. Cultural capital is knowing that cashmere is a more aspirational fabric than cotton or that a Jackson Pollock painting is much more than a mess of drips that a child could replicate. (Bourdieu points out that, in the West, being open to radical aesthetic experimentation and abstraction are markers of the elite class.) Not only is it an understanding of art; it is an understanding of what art means in a social context, and what different pieces or artists symbolize. Content capital, then, is fluency in digital content: the knowledge of what kinds of content to produce, how the feeds of various platforms work, what they prioritize, and how audiences might react to a given creation. Those who have more content capital gain more followers, and thus more power in the cultural ecosystem of Filterworld.

这条规则是网络生活的一个平凡事实，它如此普遍以至于我们很容易忘记它的存在：更多的粉丝和更多的参与度总是被认为更好。（人们不应该再满足于仅仅拥有数百名粉丝，尽管这可能是你在现实世界中认识的人的极限。）主要的动机是让粉丝数量增加。艾希霍恩写道：“人们只需在网上闲逛，更准确地说，通过发布引起反响的内容，就能积累自己的内容资本，进而带来更多的粉丝和更多的内容。”内容资本可以用来转化为其他形式的资本：通过出售赞助或直接向粉丝销售产品来赚钱，无论是印有影响者账户的 T 恤还是印有作者名字的精装书。这两者并没有太大的区别，都是特定人粉丝群的物理象征，也称为swag。

The rule is a mundane fact of online life, so pervasive that it’s easy to forget it exists: more followers and more engagement are always posed as better. (One is no longer supposed to be satisfied with mere hundreds of followers, though that might be the limit of people you know in the physical world.) The primary incentive is to make the numbers go up. “One builds up one’s content capital simply by hanging out online and, more precisely, by posting content that garners a response and, in turn, leads to more followers and more content,” Eichhorn wrote. That content capital can be used or transformed into other forms of capital: making money by selling sponsorships or selling products directly to followers, whether a T-shirt emblazoned with an influencer’s account or a hardcover book with an author’s name. The two are not so different as physical symbols of fandom for a particular person, also known as swag.

这个等式适用于任何文化产品或领域：粉丝越多，赚的钱就越多。社交媒体将文化量化为一套陈词滥调的指标，衡量浏览量、点击率，以及最终的购买率。最终的结果是同质化，因为创作者都追逐同样的动机来吸引更多关注，复制在特定时刻最有效的模式。“随着时间的推移，不同媒介（例如电影、视频、音频）之间的区别逐渐消失。”“内容（录音、印刷书籍）以及类型（例如，非虚构与虚构，电视剧与情景喜剧）之间的竞争似乎不那么重要，”艾希霍恩写道。她这样描述这场无休止的竞争：“越来越重要的是，人们在生产内容，并且频率和数量越来越高。” 在书中的其他地方，艾希霍恩的表述更加简单直接：“内容产生内容。”

This equation holds for any cultural production or field: The more followers, the more money you make. Social media has quantified culture into a banal set of metrics measuring views, click-throughs, and, ultimately, purchase rates. The net effect is homogenization as creators all chase the same incentives to attract more attention, copying whichever formula works best in a given moment. “As time goes on, the distinctions between mediums (e.g., film, video, audio recording, printed book) and between genres (e.g., nonfiction versus fiction or television drama versus sitcom) seem to matter less,” Eichhorn wrote. She described that endless race: “Increasingly, what matters is simply that one is producing content and doing so at an increasingly high frequency and volume.” Elsewhere in the book, Eichhorn puts it more simply and brutally: “Content begets content.”

这并不是说内容催生艺术。事实上，算法推送所要求的过剩内容往往妨碍了艺术创作，因为它占用了创作者越来越多的时间。“过去专注于写书、拍电影或创作艺术的文化生产者，现在也必须花费大量时间制作（或付费请人制作）关于他们自己和他们作品的内容，”艾希霍恩写道。任何尝试过在网上开展事业的人——比如烘焙义卖、派对、艺术装置——都知道，辅助内容很快就会分散注意力。作者（我）忙着在 Instagram 上晒他那杂乱无章的艺术办公桌，宣传他的作家身份，以及查看后续点赞，根本无暇真正写书。

That is not to say that content begets art. In fact, the excess content demanded by algorithmic feeds more often gets in the way of art, because it sucks up an increasingly high percentage of a creator’s time. “Cultural producers who, in the past, may have focused on writing books or producing films or making art must now also spend considerable time producing (or paying someone else to produce) content about themselves and their work,” Eichhorn wrote. As anyone who has tried to get an undertaking off the ground online—a bake sale, a party, an art installation—knows, ancillary content quickly becomes a distraction. The author (me) is too busy Instagramming his artfully cluttered desk, broadcasting his writerly identity, and checking for subsequent likes to actually write his book.

我们每个人都必须关注自己的网络内容资本，即使只是和朋友交流。某些餐厅或度假胜地比其他餐厅拥有更高的内容资本，从而带来更优质的Instagram帖子，就像某些人生里程碑在Facebook动态中更引人注目一样。人们可能会利用生日或婚礼来获得一些额外的宣传。但这种曝光并不总是对个人有利。

We all have to worry over our content capital online, even if we’re only communicating with our friends. Certain restaurant meals or vacation destinations have more content capital than others, resulting in better Instagram posts, just as certain life milestones make more of a splash in Facebook feeds. One might take advantage of a birthday or wedding to get some extra promotion. But the exposure is not always personally affirming.

在TikTok上，即使视频内容是讲述个人创伤或展示新的艺术创作，评论区也会充斥着疯狂的提问，比如创作者穿什么牌子的衣服，家里有什么家具，仿佛这些推荐就是他们最引人注目的文化贡献。我记得有一位女士在TikTok上对着她的iPhone摄像头，讨论了一次职场骚扰事件。评论者们追问她的露背上衣是哪家的；品牌是Zara，但这当然不是视频的重点。甚至连个人脆弱性的表达也被降格为视觉内容。我观察到一位专注于美食的创作者，除了分享她烹饪过程（切菜炒菜做汤）的平和视频外，还分享了她遭受种族歧视的经历。这种视觉氛围被用来掩盖个人内容，或许是为了更容易地吸引那些在算法推送中分心的观众。

On TikTok, even when a video might show a person recounting a personal trauma or displaying a new artistic creation, the comments are full of rabid questions about which brands of clothing the creator is wearing or the furniture that they have in their home, as if those recommendations would be their most compelling cultural contribution. I recall one woman on TikTok discussing an incident of harassment at work, facing the camera of her iPhone. Commenters badgered her for where her halter top was from; the brand was Zara, but that certainly wasn’t the point of the video. Even the expression of personal vulnerability was lowered to the level of visual content. One food-focused creator I observed took to discussing her experiences of racism on top of placid videos of her cooking processes, chopping and sautéing vegetables to make a soup. The visual ambience was used to mask the personal content, perhaps making it more likely to reach the distracted viewer within the algorithmic feed.

构建内容资本会分散注意力，尽管这是接触消费者的必要手段。在 Filterworld，创作艺术却不以消费为目标几乎是不可想象的。这与卡门·埃雷拉 (Carmen Herrera) 这样的画家截然不同，她直到九十多岁才声名鹊起，并于 2022 年去世，在近一个世纪的时间里默默无闻地创作着极简主义的画布。如今，我们常常会觉得，没有关注就没有创造力，而没有算法推荐的加速器，关注就无法产生。

Building content capital is a distraction, albeit a necessary one to reach consumers. And in Filterworld, making art without the goal of it being consumed is almost unimaginable. It’s the opposite of the way a painter like Carmen Herrera, who didn’t achieve fame until her nineties and passed away in 2022, toiled in obscurity on her minimalist canvases for the better part of a century. Today, it can often feel like there is no creativity without attention, and no attention without the accelerant of algorithmic recommendations.

我的个人内容资本只是中等水平。我从事职业记者工作十多年，而且从2008年Twitter成立之初就一直使用Twitter，但在这个平台上，我“仅仅”拥有大约两万六千名粉丝。在Instagram上，我只有四千多一点——这个数字我也说不清道不明，因为我发布的内容大多只是一些现实生活中的日常琐事：我做的晚餐、我家狗狗的照片，以及偶尔模仿我最喜欢的艺术摄影师之一斯蒂芬·肖尔（Stephen Shore）的静态照片和普通场景。我发布的图片主要供自己欣赏，构建一个几乎不面向大众的个人档案。因此，它们并没有引起太多的互动。

My personal content capital is only middling. Through more than a decade of working as a professional journalist and being on Twitter since its early days in 2008, I “only” have around twenty-six thousand followers on the platform. On Instagram, I have just over four thousand—a number I can’t quite explain, since I don’t post much more than mundane evidence of my IRL existence: dinners I cook, photos of my dog, aggressively static and normal scenes that sometimes mimic Stephen Shore, one of my favorite art photographers. The images I post are largely for my own consumption, building a personal archive barely directed at a wider audience. As such, they don’t command much engagement.

我很久以前就决定不让自己的声音完全适应算法推送——或者与其说是决定，不如说是觉得自己做不到。我不够酷，不够风趣，不够有魅力，也不够权谋，所以不知道该如何操控哪些变量。我的推文很少爆红，我的生活方式也绝对不足以填满Instagram推送。我永远也成不了那样的网红。然而，这些年来，我意识到，即使我不是拥有数十万粉丝的Instagram网红，我对一小群关注我发布内容的人还是有影响力的，最终他们的关注确实会影响我发布的内容。

I decided long ago against fully adapting my voice to the algorithmic feed—or perhaps not decided so much as felt incapable of doing it. I wasn’t cool enough, funny enough, attractive enough, or Machiavellian enough to know exactly which variables to manipulate. My tweets rarely went viral, and I certainly didn’t have enough lifestyle material to fill an Instagram feed. I could never be that kind of influencer. Yet over the years, I realized that even though I wasn’t an Instagram star with hundreds of thousands of followers, I was influential to a set group of people who were watching what I was posting, and eventually their attention did influence what I posted.

我发现有一些方法可以展示我正在做的事情，从而最大化我的内容资本。我费尽心思发推文分享我的最新文章，试图找到合适的方法。被分享最多的内容：一个引人好奇的标题，留下一个开放性的问题，或者突出故事中最戏剧性的引言。我迎合我的在线受众，触及人们熟知的话题：小众设计评论、对硅谷文化的抱怨、艺术史参考。我个人生活的证据，比如对我工作的咖啡店的抱怨，总是很受欢迎。问题是，我开始将推送内容奖励的主题与我的个人品味混淆——我写推特想看的内容，这开始遮蔽我对自己会写什么或感兴趣什么的认识。

I found that there were certain ways I could present the things I was doing to maximize my possible content capital. I labored over tweets to share my latest articles, trying to figure out what would get shared the most: a curiosity-gap headline that left a question open, perhaps, or highlighting the most dramatic quote in a story. I catered to my online audience, hitting the subjects that people knew me for: niche design criticism, complaints about Silicon Valley culture, art-historical references. Evidence of my personal life, like a complaint about a coffee shop I was working in, always did well. The problem was that I began to confuse the subjects rewarded by the feed with my personal taste—I wrote what Twitter wanted to see, which began to occlude my awareness of what I would have written or been interested in on my own.

帖子瞬间获得的点赞浪潮不仅有助于推荐算法评估哪些内容值得推广，也为创作者提供了前所未有的实时衡量标准，让他们了解哪些内容能引起受众的共鸣，就像每个想法都能通过智能手机上的尼尔森收视率来衡量一样。文化扁平化是其后果之一。但同样的机制也使我们的公共政治话语越来越极端，因为冲突和争议点亮了信息流，并以微妙和模糊性永远无法企及的方式吸引点赞。

The instantaneous wave of likes in response to a post helps the recommendation algorithm evaluate which pieces of content should be promoted, but it also gives the creator an unprecedented real-time measure of what resonates with their audiences, as if every thought was gauged by a Nielsen rating viewable on a smartphone. Cultural flattening is one consequence. But the same mechanism is also what makes our public political discourse more and more extreme, because conflict and controversy light up the feed and attract likes in a way that subtlety and ambiguity never will.

即时诗歌

INSTA-POETRY

数字平台和算法推送所奖励的文化类型，无论在形式还是内容上，都与传统人类引领潮流的模式所取得的成功截然不同。诗歌是算法推送改变艺术形式及其公众接受度的最明显例子之一。过去十年，Instagram 上涌现出一代“Instagram 诗人”，他们根据平台的结构和需求调整作品，向粉丝售出了数百万册书籍。其中最著名的一位是鲁皮·考尔 (Rupi Kaur)：这位 1992 年出生于印度的加拿大女性，通过发布简短的诗歌（这些诗歌被分成几行，与她自己创作的简单线条画一起融入 Instagram 的图片框中），积累了 450 万粉丝。这些诗歌采用标准衬线字体，全部小写，如同在手机键盘上匆匆写下——尽管也参考了旁遮普语，这种语言没有大小写之分。其中一首诗的部分内容是“他们没有告诉我这会让我如此痛苦 / 不有人警告我/与朋友相处时会经历心碎。” 这些诗歌图像的署名是“-rupi kaur”，这一举动强化了个人作者身份——一个品牌——同时也充当了一种作者水印，以防图像（很有可能）脱离其原始数字语境或被转发到其他地方。观看者始终知道作者是谁。

The kinds of culture rewarded by digital platforms and algorithmic feeds are markedly different from what was successful in the traditional model of human tastemakers, both in form and content. Poetry offers one of the most blatant examples of algorithmic feeds altering both an art form and its public reception. Over the past decade, a generation of “Insta-poets” have emerged on Instagram and sold millions of books to their followers by shaping their work to the structure and demands of the platform. Rupi Kaur is the most famous: a Canadian woman born in India in 1992, she has amassed 4.5 million followers by posting brief poems, broken into short lines, fit into the Instagram image square along with her own simple line drawings. The poems are written in a standard serif font and all lowercase, as if dashed off on a phone keyboard—though also in reference to the Punjabi language, which doesn’t have different cases. One reads, in part, “they did not tell me it would hurt like this / no one warned me / about the heartbreak we experience with friends.” The poem-images are signed “- rupi kaur,” a gesture that reinforces personal authorship—a name brand—but also serves as a kind of authorial watermark, should the image, as is very likely, become divorced from its original digital context or get reposted elsewhere in the feed. The viewer always knows who wrote it.

其他 Instagram 诗人是男性，他们拥护一种自觉脆弱的刻板男性气质。其中一位名叫 Atticus，是一位匿名诗人，在 Instagram 上拥有 160 万粉丝，著有多本畅销书。Atticus 的诗歌经常出现在酒瓶或度假人士的库存照片上，关注爱情、美和醉酒等主题：“亲爱的 / 让你和我一起闲逛一会儿”是一篇完整的帖子。还有拥有 280 万粉丝的 RM Drake，他更喜欢用散文诗的形式来表达自我激励的箴言：“如果你失去了一个对你毫不在乎的人，那也不算损失，”一首短诗的开头写道。RM Drake 也将自己的名字添加到每一帧 Instagram 画面中，以 Kaur 的兄弟版出现。他们共同发展了 Instagram 诗歌的通用风格，与国际极简主义咖啡馆有着同样的同质性。其风格遵循了我一直概括的 Filterworld 的原则：诗歌既要兼具图像功能，又要兼具文本功能，并能在各种数字平台上流畅传播，无论是 Instagram 动态、Facebook 帖子还是 TikTok 幻灯片。诗歌本身的内容必须具有可关联性和可分享性，更少地诉诸个人的经历或视角，而更多地探讨普遍的、可识别的主题。

Other Insta-poets are men espousing a self-consciously vulnerable form of stereotypical masculinity. One goes by the name Atticus, a pseudonymous poet with 1.6 million Instagram followers and multiple bestselling books. Atticus’s poems are often splashed across stock photos of wine bottles or people on vacation, focusing on themes of love, beauty, and getting drunk: “My darling / let’s you and I ramble on this life a while” is one post in its entirety. Then there’s R. M. Drake, 2.8 million followers, who prefers prose-poem paragraphs that echo rote self-help mantras: “It is not a loss if you lost someone who didn’t give a fuck about you,” one brief poem begins. R. M. Drake, too, appends his name to every Instagram frame, appearing as a bro version of Kaur. Together they have developed the generic style of Insta-poetry, with the same homogeneity as the international minimalist coffee shops. The style follows the principles that I’ve been outlining as characteristic of Filterworld: The poems have to function as images as much as text and travel seamlessly through various digital platforms, whether Instagram feed, Facebook post, or TikTok slide. The content of the poems themselves has to be relatable and sharable, speaking less to an individualized experience or perspective and more to universal, recognizable themes.

考尔既是这场运动的代言人，又成了替罪羊，这多少有些不公平。然而，这个平台本身或许才是造成“即时诗歌”空洞审美的罪魁祸首。考尔十几岁时就开始朗诵诗歌，之后转向Tumblr，2013年又在Instagram上发表作品。2014年，她还自费出版了第一本诗集《牛奶与蜂蜜》。随着她在网上名气渐长，这本书于2015年由一家专业出版商重新发行，销量超过200万册，并荣登《纽约时报》畅销书排行榜榜首。然而，长期以来一直是诗歌这一小众艺术形式唯一评判者的评论家们，大多对考尔的作品嗤之以鼻。他们把考尔比作把它抄成贺卡，并抱怨它的字面意思。它不值得长时间思考。

Somewhat unfairly, Kaur has been made both the face and the scapegoat of the movement, when the platform itself may be more to blame for Insta-poetry’s vapid aesthetics. Kaur first performed spoken-word poetry as a teenager, then moved to posting on Tumblr, and then Instagram in 2013. In 2014, she also self-published her first collection, Milk and Honey. As her online fame grew, the volume was re-rereleased by a professional publisher in 2015 and went on to sell over two million copies and hit number one on the New York Times bestseller list. Critics—long the sole arbiters of the niche art form of poetry—largely hated Kaur’s work, however. They likened it to greeting card copy and bemoaned its literalness. It didn’t bear thinking about for any length of time.

一些评论或许无意中反映出她的诗歌在多大程度上满足了Instagram的传播需求：“考尔的诗歌表达了显而易见、略带趣味的意识流，并以视觉上引人入胜的方式展现思想，”评论家法雷亚·菲苏丁（Fareah Fysudeen）写道。“她的诗歌符合常理，浅显易懂，却又空洞无物，在没有深度的地方营造出一种错觉。” 在Instagram上，一切都必须是视觉化的：因此考尔的大量换行符才显得图形清晰，图画也增添了趣味。如果一篇帖子的设计初衷是出现在信息流中，吸引眼球几秒钟，然后毫无冒犯地滚过，再也不会被看到，那么肤浅就不是什么缺陷了。小说家鲁曼·阿拉姆 (Rumaan Alam) 在 2019 年为传统文学批评堡垒《新共和》撰写的一篇题为《鲁皮·考尔是十年来最杰出的作家》的文章中写道：“考尔作为一名艺术家的成就在于，她的作品在形式上体现了定义当代生活的技术：智能手机和互联网。”

Some of the reviews perhaps inadvertently reflected how much her poems answered the demands of Instagram for distribution: “Kaur’s poetry states obvious, mildly interesting stream-of-consciousness shower thoughts in visually appealing ways,” one critic, Fareah Fysudeen, wrote. “Her poems are expected, obvious, and vacuous, painting an illusion of depth where there is none.” On Instagram, everything must be visual: hence the graphic clarity of Kaur’s plentiful line breaks and the added interest of the drawings. Superficiality is not such a flaw when a post is designed to emerge in the feed once, attract the eyes for a few seconds, and then scroll by without offense, never to be seen again. “Kaur’s achievement as an artist is the extent to which her work embodies, formally, the technology that defines contemporary life: smartphones and the internet,” the novelist Rumaan Alam wrote in a 2019 essay for The New Republic, a bastion of traditional literary criticism, titled “Rupi Kaur Is the Writer of the Decade.”

2023年，我打电话给阿拉姆，想了解他对当时自己判断的看法，当时他的判断引发了其他评论家的强烈反对。“我觉得自己被证明是正确的，”他说。“那些一眼就能看完、根本不用移动手机屏幕的文字更有吸引力。”他以玛吉·史密斯2016年的诗歌《好骨头》为例，这首诗可以压缩成一张截图，经常被当作表情包分享在推特或Instagram快拍上。“人生苦短，但我瞒着我的孩子们。/人生苦短，我缩短了我的人生，”这首诗像咒语一样开头，十五行之后就结束了。阿拉姆说，鲁皮·考尔“算不上一位伟大的诗人，但这并不意味着她没有巨大的影响力。她发现了品味形成方式和传播方式的变化。”“艺术会根据当代生活的注意力广度进行调整，”他继续说道。

I called Alam in 2023 to see how he felt about his judgment at that time, which had prompted some backlash from other critics. “I feel vindicated,” he said. “Text that can fit comfortably inside a glance, where you don’t have to move your phone screen at all, is much more appealing.” He cited the example of the 2016 poem “Good Bones” by Maggie Smith, which can be compressed into a single screenshot and is often shared as a meme, on Twitter or in Instagram stories. “Life is short, though I keep this from my children. / Life is short, and I’ve shortened mine,” it begins, like an incantation, before ending fifteen short lines later. Rupi Kaur, Alam said, “is not really a great poet, but that doesn’t mean that she’s not hugely influential. She identified something about a change in how taste is made and a change in how stuff is disseminated.” “Art will adjust to meet the attention span of contemporary life,” he continued.

考尔本人的视觉冲击力也丝毫不逊色：她修长的脸庞、棱角分明的下巴，以及一双充满戏剧感的大眼睛，都让她的自拍照成为她Instagram账号上众多自拍照中的焦点。近年来，她逐渐形成了诗歌意象和自画像交替出现的模式，大多数她摆出各种姿势（而非标准的随意快照），身着精致的高级时装，化着精致的妆容。她将自己的形象与艺术作品融为一体，这种融合并非每位创作者都能做到或渴望做到。（我们每个人对美的评价都是不平等的，这其实是另一种算法。）一个人必须拥有吸引力才能成为一名成功的艺术家吗？当然不是，但这确实有帮助，尤其是在Instagram时代。

It doesn’t hurt that Kaur herself is visually striking: her long face, sharp jawline, and wide, dramatic eyes star in the copious selfies that also fill her Instagram account. In recent years, she has settled into a pattern of alternating between poem-images and self-portraits, most fully posed (rather than the standard casual snapshots), costumed in elaborate high-fashion outfits and makeup. She poses her physical presentation as equal to her art, a conflation that not every creator is capable of or desires to make. (We are all unequal in the very human evaluation of beauty, which is a different kind of algorithm.) Does one have to be attractive to be a successful artist? Of course not, but it helps, particularly in the Instagram era.

考尔和她的诗歌拥有极高的内容资本。她既是诗人，也是影响者，这两种身份在她身上并不互相抵消，反而相辅相成：她之所以成为受欢迎的影响者，部分原因在于她的诗歌，而她的诗歌之所以受欢迎，部分原因在于她建立的在线平台。同样的情况也发生在摄影师帕特里克·贾内尔身上——我们忽略了他的艺术造诣，而更关注他的生活情色片——或许也发生在小说家肖恩·索尔·康罗身上。他2022年出版的《Fuccboi》采用了社交媒体和短信的语言，带有简洁的格言和数字俚语。康罗的个人Instagram账户上充斥着忧郁的吸烟自拍照，这些照片与他小说的最终封面很相似，记录并戏剧化地展现了这位以坏男孩小说家形象出现的作家的生活。作为一名职业艺术家一直意味着成为某种公众人物，但在Filterworld，完美的形象是艺术的先决条件。

Kaur and her poetry possess a maximum of content capital. She is an influencer as much as a poet, two identities that in her case don’t cancel each other out but are mutually reinforcing: she is a popular influencer in part because of her poetry, and her poetry is popular in part because of her established online platform. The same dynamic is true for Patrick Janelle as a photographer—we overlook his artistry in favor of his lifestyle porn—and perhaps a novelist like Sean Thor Conroe, whose 2022 Fuccboi adopted the language of social media and text messages, with clipped aphoristic lines and digital slang. Conroe’s personal Instagram account is filled with moody cigarette-smoking selfies that resemble his novel’s eventual cover, documenting and dramatizing the life of the writer in the mold of a bad-boy novelist. Being a working artist has always meant being something of a public figure, but in Filterworld the perfected image is a prerequisite to the art.

考尔意识到Instagram如何对她的作品施加审美压力，使某些主题比其他主题更受关注。在2017年接受《娱乐周刊》采访时，她表示：“我在网上最喜欢爱情诗和心痛主题的作品。” 例如，关于性暴力的诗歌获得的点赞较少。她一度尝试只发布关于这些基本主题的帖子，持续了一个月，参与度有所上升。但她表示，这感觉并不真诚。艺术家必须积极抵制社交媒体的诱惑，抵制那种总是想给观众更多他们点击率最高的内容的诱惑（记住，观众在做决定时也会被信息流操纵）。“当你沉浸其中时，你会让别人的想法改变你想要创作的内容和写作的内容，”考尔说。持续被动地消费社交媒体导致人们缺乏深刻的自我意识，考尔承认这一点，尽管她的作品利用它来装点她把这些陈词滥调当成了独到的见解：“我们太沉迷于网络，以至于几乎忘记了自我，”她说。根据一些采访，考尔的手机上没有社交媒体，而且像许多名人一样，她有一支团队来管理她的实体账户，这让她与网络形象保持距离，尽管这暗示着亲密感，但她也因此受益。

Kaur is aware of how Instagram exerts aesthetic pressure on her work, rewarding some subjects with more attention than others. In a 2017 interview with Entertainment Weekly, she said, “I get the most love online for love poems and heartache.” Poems about sexual violence, for example, got fewer likes. At one point, she switched to only posting about those basic subjects for a month, and her engagement numbers went up. But it also didn’t feel genuine, she said. The artist must work actively against the incentives of social media, the constant temptation to just give audiences more of what they click the most (keeping in mind the fact that audiences are equally manipulated by the feed in their own decisions). “When you dive too deep, you let people’s ideas change what you want to create and what you want to write,” Kaur said. The constant passive consumption of social media has led to a lack of deep self-awareness, which Kaur acknowledges even as her work takes advantage of it to dress platitudes up as singular insight: “We’re so plugged in that we’re almost not plugged into ourselves,” she said. According to some interviews, Kaur doesn’t have social media on her phone and, like many celebrities, has a team of staff to manage her actual accounts, separating herself from her online presence even while benefiting from its suggestion of intimacy.

任何将社交媒体视为艺术对立面的评价都带有精英主义的色彩。并非每个人都能接触到传统的、更容易被接受的艺术创作途径：常春藤盟校、文学杂志、切尔西画廊。利用引人注目的图像或网络影响力，可以培养最初的受众，证明艺术家无需展现的既定兴趣，并打破封闭的生态系统。读者可能会被艺术家在网络上的可及性以及艺术家作为影响者的影响力所吸引。例如，数百万人对考尔的作品非常欣赏，以至于他们购买她的书籍并在Instagram上关注她，而她的每一幅诗歌图片都获得了数十万个赞。事实上，近年来诗歌消费总体呈增长趋势，Poets.org 的访问量从 2020 年到 2021 年增长了 25%，诗集销量也大幅提升——在英国，2018 年至 2019 年的诗歌销量增长了 12%。这或许部分归因于互联网鼓励了人们消费文本片段，但也可能是过去几年的混乱激起了人们对精神沉思的渴望。文学诗歌——或者说被精英机构认可的诗歌——仍然很容易找到，与 Instagram 诗歌并存，尽管销量通常要低得多。

There is an element of elitism at play in any evaluation that casts social media as the opposite of art. Not everyone has access to the traditional, more acceptable routes of art making: Ivy League universities, literary magazines, Chelsea galleries. Leveraging a compelling image or Internet presence can be a way to cultivate an initial audience, prove a level of established interest that some artists don’t have to demonstrate, and break into the closed ecosystem. Readers can be attracted to the accessibility of the figure of the artist online, the artist-as-influencer. Millions of people value Kaur’s work, for example, enough that they buy her books and follow her on Instagram, and her poem-images each rack up hundreds of thousands of likes. In fact, poetry consumption overall has increased in recent years, with traffic to Poets.org up 25 percent from 2020 to 2021 and higher sales numbers for poetry books—in the United Kingdom, poetry sales increased 12 percent from 2018 to 2019. It might be due in part to how the Internet has encouraged the consumption of fragments of text, though it could equally be the chaos of the last several years inciting a desire for spiritual contemplation. Literary poetry—or poetry that is accepted by the establishment of elite institutions—is still easy to find, too, coexisting with Insta-poetry, though usually with much lower sales figures.

在文学领域，我欣赏1957年出生的韩裔美国诗人金明美（Myung Mi Kim）的诗歌。金的一些诗歌或许同样适合用Instagram图片来展现，但它们远非一眼就能看懂。她的诗《土地的积累》由三列短语构成——“清点牲畜的财产”、“生产继承人的数量”、“生育子女的抚养”——可以以任何顺序动态地阅读。这些短语唤起了一种古老而又人类史前时代的感觉，但并未将其具体化。她诗歌中强烈的碎片化，恰恰反映了我们自身语言的碎片化。时代。在另一首短诗（尽管可能无限长）《[引言：‘名字以何种方式’]》中，她用几行字就提出了一个宏大的解构：“名字以何种方式被赋予事物。过滤。并非每个被运用过的词语都还存在。”（第一句话的过去时态起初很微妙，随后却令人心碎。）

In literary context, I appreciate the poetry of Myung Mi Kim, a Korean American poet born in 1957. Some of Kim’s poems might equally fit in an Instagram image, but they are far less immediately scrutable. Her poem “[accumulation of land]” is a grid of phrases in three columns—“counting herds possessions,” “production heirs number,” “bearing child rearing”—that are meant to be read dynamically in any order. The phrases evoke, but do not pin down, a feeling of antiquity and human prehistory. The aggressive fragmentation of her poetry speaks to the very fragmentation of language in our own era. In another short (though perhaps infinite) poem “[Exordium: ‘In what way names’],” she suggested a vast deconstruction in a few lines: “In what way names were applied to things. Filtration. Not every word that has been applied, still exists.” (The past tense of the first sentence is at first subtle and then heartbreaking.)

清晰明了、简洁易懂的字面意思，与语言上的困难和不得不接受的优柔寡断形成了鲜明对比：一种美学方法本身并无优劣之分；它们只是不同的选择。然而，在“过滤世界”（Filterworld）中，我们面临的文化环境不可避免地优先考虑前者，因为它通过算法信息流传播得更有效率，而除了这些信息流之外，创作者能够接触到在如此资本主义环境下生存所需的受众的渠道越来越少。无论是在推特、Instagram、抖音还是亚马逊，艺术家都在与这股无形的力量抗争。他们可以虔诚地追随个人愿景，或许会因此牺牲参与度和未来的谋生潜力；或者，他们可以同时迎合信息流和受众——创作出相当于只写爱情和心碎的诗歌——希望能够以网红的身份生存下去，如果艺术本身无法盈利，他们还可以通过出售赞助或赠品来维持运营。最终，算法推送可能不会导致艺术的消亡，但它往往会对艺术的发展造成阻碍。

Blatant clarity and simple, literal takeaways versus linguistic difficulty and the need to accept irresolution: one aesthetic approach is not better or worse than the other; they are simply different sets of choices. Yet in Filterworld, we face a cultural environment that inevitably prioritizes the former over the latter because it travels more effectively through algorithmic feeds, and there are fewer and fewer outlets outside of those feeds available for creators to access the audiences they need to survive in such a capitalistic environment. Whether Twitter, Instagram, TikTok, or Amazon, the artist contends with the invisible force. They can devoutly follow a personal vision, perhaps to the detriment of their engagement numbers and future potential to make a living, or they can cater to both feeds and audiences—creating the equivalent of only love and heartbreak poems—in the hope of surviving as an influencer, with the backup of selling sponsorships or swag if the art itself doesn’t profit. Ultimately, the algorithmic feed may not be the death of art, but it often presents an impediment to it.

BOOKSTARAM 和 BOOKTOK

BOOKSTAGRAM AND BOOKTOK

社交媒体网红是市场营销的强大力量：他们几乎可以推广任何东西，无论是时尚品牌的新产品，还是具有破坏性的政治意识形态。如果你想让某样东西在 Filterworld 上流行起来，最快的方法就是让网红站在你这边。整个行业都围绕着他们的引力被重塑。汉娜·奥利弗·德普 (Hannah Oliver Depp) 决定在家乡华盛顿特区开设一家名为 Loyalty Books 的新书店公司（她于 2018 年启动了这个项目），她决定围绕“Bookstagram”这个蓬勃发展的网红群体来设计公司。这些网红并非专注于时尚或旅游，而是通过推荐书籍来吸引受众，他们在沉稳的出版业中的影响力和权威正在不断增强。“对我来说，这是一个机会”“创造一个我愿意待在的美好空间，我知道那正是Instagram发布内容所需要的空间，”德普告诉我。她推出了一些快闪书店，里面摆放着一张风景如画的蓬松扶手椅，顾客可以坐在里面摆姿势拍照。后来，她还将自己的书店改造成TikTok的文学网红平台Booktok，为拍摄视频而非静态照片腾出了空间。

Social media influencers are forces of marketing: They can popularize almost anything, whether a new product from a fashion brand or a damaging political ideology. If you want something to be popular in Filterworld, the fastest way is to get the influencers on your side. Entire industries have been reshaped around their gravity. When Hannah Oliver Depp decided to open a new bookstore company called Loyalty Books in her hometown of Washington, D.C., a process she began in 2018, she decided to design it around the burgeoning group of influencers that made up “Bookstagram.” Rather than fashion or travel, these influencers gained their audiences by recommending books, and their power and authority in the staid publishing industry was growing. “There was an opportunity for me to create a beautiful space that I would want to be in, knowing that was the kind of space Instagram needed for content,” Depp told me. She launched pop-up bookshops with a picturesque poofy armchair that visitors could pose for photos in. Later, she adapted her stores to the literary-influencer side of TikTok, called Booktok, too, making space for shooting videos instead of still images.

我和德普的初次相识是通过一位我们共同的小说家朋友；她当时正来华盛顿进行新书巡回宣传，邀请我在德普位于华盛顿特区的商店举办她的活动。由于疫情持续不断且难以预测，我和这位小说家虽然见面了，但我们的对话通过Zoom视频会议系统传输给了不在场的观众。这超现实时刻的一大好处是，活动结束后，我和小说家德普一起去了德普最喜欢的酒吧——红德比（Red Derby）。这家酒吧杂乱无章，墙壁凌乱，宽敞的屋顶酒吧供应丰富的威士忌姜汁。红德比以其地理位置和社区为傲：它不刻意迎合Instagram，也不刻意在Yelp上博取好评。这家酒吧及其老板以善于为当地创意人士提供优质服务而闻名，他们营造了一种热情好客的氛围，这也激励了德普发展自己的事业。

I first met Depp through a mutual novelist friend; she was coming through the city for her book tour, and she asked me to host her event at Depp’s D.C. shop. Given the endless and unpredictable surges of the pandemic, the novelist and I were together in person, but our conversation was beamed out over Zoom to a physically absent audience. The upside to the surreal moment was that after the event Depp, the novelist, and I headed to Depp’s favorite bar, Red Derby. It’s a rambling, ramshackle place with cluttered walls and a wide rooftop bar that offers plentiful whiskey gingers. Red Derby is the kind of bar that is proudly of and for its geographical place and community: It doesn’t try to cater to Instagram or cultivate positive Yelp reviews. The bar and its owner are known as a great employer for the local creative set, building a sense of hospitality that was also an inspiration for Depp in growing her own business.

在Red Derby酒吧，我和德普边喝边聊互联网对书籍的影响：创作书籍、销售书籍、寻找书籍以及阅读书籍。德普天生就具有权威性；作为一名黑人酷儿女性，在白人至上、同质化且古板的图书出版和销售领域，她必须开辟自己的道路，建立一种新的机构。她最初学习的是艺术史，这让她对视觉文化产生了兴趣，并在华盛顿特区的“政治与散文”等知名书店工作过。在推特上，后来又在Instagram上发现了一个由其他有色人种组成的社群，这让她更有信心独立创业。（算法推送也能帮助人们找到彼此，建立社群。）忠诚度最初没有固定的地点，但德普精心布置了她的临时店面装置，使其具有永久性。她模仿了自己家客厅的审美：“古董店与宜家的结合，”她说。

Over various drinks at Red Derby, Depp and I discussed the impact of the Internet on books: writing them, selling them, finding them, and reading them. Depp is a naturally authoritative presence; as a Black, queer woman in the very white, homogenous, and staid world of book publishing and selling, she has had to chart her own path and build a new kind of institution. She originally studied art history, which gave her a taste for visual culture, and worked in established bookstores like D.C.’s Politics and Prose. Discovering a community of other people of color talking about books first on Twitter and then on Instagram gave her more confidence to venture out on her own. (Algorithmic feeds can also help people find each other and build communities.) Loyalty began without a fixed location, but Depp was careful to arrange her temporary storefront installations with a sense of permanence. She copied the aesthetic of her own home living room: “Antiques store meets IKEA,” she said.

她注意到华盛顿的书籍博主非常集中，部分原因是该市的初级产业，雇佣受过良好教育、有空闲时间来维护社交媒体账户的官员。书评博主也感到与文学界隔绝。他们被忽视，而纸质出版物则更青睐传统的评论家；直到最近几年，出版商才开始直接与有影响力的人合作，分发难以获得的样稿。德普说，书评博主“往往是女性或酷儿”。“她们想要一家不会问‘你为什么要拍照？’或将她们的阅读品味视为低劣的书店。”通过社交媒体走红的书籍通常是浪漫小说和奇幻小说，比如凯西·麦奎斯顿的《红、白与皇家蓝》，讲述了一个美国人和威尔士亲王的同性恋爱情故事，以及VE·施瓦布的《艾迪·拉鲁的隐形生活》，讲述了时间旅行和永生的诅咒。

She had noticed that Washington had a high concentration of bookstagrammers, in part due to the city’s primary industries, which employ well-educated bureaucrats with the free time on their hands needed to maintain social media accounts. Bookstagrammers also felt isolated from the literary establishment. They were ignored in favor of traditional critics at print publications; publishers started working directly with influencers only in the last few years, handing out hard-to-get galleys. Bookstagrammers “tend to be female or queer,” Depp said. “They wanted a bookstore that wouldn’t be like, ‘Why are you taking pictures?’ or treat their reading taste as inherently lesser.” The kinds of books that become famous via social media were often romances and fantasy novels, like Casey McQuiston’s Red, White & Royal Blue, a gay love story about an American and the Prince of Wales, and V. E. Schwab’s The Invisible Life of Addie LaRue, about time travel and a curse of immortality.

所以，当德普搬进她长期经营的店铺时，这些空间必须充满引人入胜的瞬间，成为绝佳的拍照背景。“视觉元素才是卖点。什么才能让人们走进来，说：‘天哪，我一定要发这个！’”德普说道。“所有东西都经过精心布置，拍下来都会很赏心悦目。我不仅要想象商店的每一寸空间，无论是在三维的现实空间中，还是在正方形或长视频中，它都能吸引你的眼球吗？” 它针对在线内容创作进行了优化，因为商店和书籍必须通过算法推送进行分发，就像在实体空间中一样：面向在线路人，也面向从人行道走进来的人。

So when Depp moved into her long-term locations, the spaces had to be full of moments of interest that could provide good photo backdrops. “It’s the visual part that sells. What would make someone come in and say, ‘Oh my god I have to post this’?” Depp said. “Everything is arranged in a way that if you take a snapshot of it, it will look good. I’m imagining every inch of the store not just in three-dimensional, real-life space, but also if it’s in a square, or a long video, does it capture your eye?” It’s optimized for online content creation, because the store as well as the books must be distributed through algorithmic feeds just as much as function in physical space: for the online passerby as well as someone walking in from the sidewalk.

德普不太清楚哪些书会在 Bookstagram 或 Booktok 上走红。Instagram 速度较慢，也更本地化；书商可以密切关注客户群的流行趋势，并确保库存吸引买家的书籍。（许多书店开始摆放“你在 Instagram 上看到的书”的表格。）但 TikTok 的即时性更强，覆盖范围更广，覆盖全球受众也更广泛。“它以​​更快的方式超出你的控制范围，”德普说。“它只是一个更强大的趋势算法。”出版商常常对哪些书籍会走红感到措手不及，他们很难印刷足够的书籍并将它们运送到合适的地点，这意味着像 Loyalty 这样的商店无法库存足够的书籍来销售给顾客。这有时会导致顾客回到亚马逊这个算法恶棍，它可以接收并保留更多的副本——部分原因是出版商优先向规模大得多的公司供货。

What Depp has less of a grasp on is which books become popular on Bookstagram or Booktok. Instagram was slower and more localized; the bookseller could keep an eye on what was trending for her customer base and be sure to stock the set of books that drew buyers in. (Many bookstores began setting out tables of “Books You Saw on Instagram.”) But TikTok was more instantaneous and reached a much wider, global audience. “It gets out of your control in a much faster way,” Depp said. “It’s simply a more powerful trend algorithm.” Publishers are often caught by surprise by which books go viral, and they struggle to print enough books and ship them to the right locations, which means that stores like Loyalty can’t stock enough copies to sell to customers. That sometimes leads customers back to the algorithmic villain of Amazon, which can receive and hold on to many more copies—in part because publishers prioritize supplying the much larger company.

影响者所推广的文学类型也具有同质性，具体到那些可以通过动态消息加速传播的书籍。“最畅销的是异性恋白人女性，她们写的是一些带有某种情感色彩的书籍：自助类、爱情类或与爱情相关的，”德普解释道。Booktok 的明星科琳·胡佛可能属于后者，而玛德琳·米勒的两部希腊神话改编作品《阿喀琉斯之歌》和《喀耳刻》在被 Booktok 收购后销量均突破百万册，将古代原型的宏伟与千禧一代对爱情和人际关系的感性（尤利西斯被描绘成一个被遗弃的花花公子）融合在一起。

There’s a homogeneity to the kind of literature that influencers promote, too, narrowing down to the kinds of books that can accelerate through feeds. “The top-selling folks are straight white women writing somehow emotional books: self-help-oriented books, romance, or romance-adjacent,” Depp explained. The Booktok star Colleen Hoover might fall into the latter category, while Madeline Miller’s two retellings of Greek mythology, Song of Achilles and Circe, which have both passed a million copies after being taken up by Booktok, mingle the grandeur of ancient archetypes with a very millennial sensibility of love and relationships (Ulysses as marooned fuckboy).

TikTok 时代的流行往往是全有或全无的，当一本书或一个话题变得流行时，就会引来想要分一杯羹的模仿者。“同质化的问题不仅仅在于它很无聊；最令人反感或最不令人反感的内容会浮出水面，因为这样能带来点击量，”德普说。“这就是本周在 TikTok 上取得成功的人的问题所在：从未读过这本书的人会制作关于它的视频，因为这是热门话题。事情一开始是出于真正的兴趣，但到了第一千个关于它的视频时，它就与这本书本身无关了。” 算法推送将书籍的表面符号与其作为文学作品的实际价值割裂开来。

TikTok-era popularity tends to be all or nothing, and when one book or topic becomes popular, it drives copycats who want to get in on the traffic. “The problems of homogeneity are not just that it is boring; the most or least offensive stuff rises to the top, because that gets clicks,” Depp said. “This is the issue about whoever is succeeding on TikTok this week: People who have never read the book are going to make a video about it, because that’s the trending topic. Things start out with genuine interest, but by the thousandth video about it, it has nothing to do with the thing itself.” The algorithmic feed alienates the superficial symbol of the book from its actual value as literature.

算法策展也可能疏远消费者。埃莉诺·斯特恩是一位作家，最近刚毕业，是 Booktok 的成员，但她也对它持怀疑态度。她于 2020 年疫情期间加入该平台，并于 2021 年初开始制作视频。我认为她是 TikTok 本土版的评论家。她对着镜头发表口头评论，内容涵盖从晦涩难懂的语言学到新书和杂志文章等各种领域。她拥有超过七万名粉丝，就她的主题而言，这是一个很高的数字。斯特恩说，就文学文化而言，TikTok 的算法推送“确实让一切都趋于千篇一律”。这是一个问题，因为“为你推荐”的推送，也就是你的个性化算法，也被视为“你潜意识的外化版本”。斯特恩观察到，TikTok 如何鼓励用户将自己融入它涵盖了特定的身份类别或类型，就像它涵盖了文化类型一样。“无论你消费什么，它都只是你自我的一种表达；它的存在只在于它能够描述你，”她说。在这个平台上，书籍的普及与其说是作为阅读的文本，不如说是作为可购买的生活方式配件，一种身份的视觉符号。这就是 Filterworld 所鼓励的自恋。

Algorithmic curation can alienate consumers, too. Eleanor Stern, a writer and recent grad student, is a part of Booktok but also skeptical of it. She joined the platform in 2020, during the pandemic, and began making videos in early 2021. I think of her as the TikTok-native version of a critic. She delivers spoken essays to the camera on everything from obscure linguistics to new books and magazine articles. She has over seventy thousand followers, a high number for her subject matter. When it comes to literary culture, TikTok’s algorithmic feed “does drive it all toward sameness,” Stern said. Which is a problem because the “For You” feed, your personalized algorithm, is also seen as an “externalized version of your subconscious.” Stern has observed how TikTok encourages users to slot themselves into particular categories or genres of identity, just as it brackets genres of culture. “Whatever it is that you’re consuming just becomes an expression of your self; it exists only insofar as it can describe you,” she said. On the platform, books are popularized less as texts to read than as purchasable lifestyle accessories, visual symbols of an identity. Such is the narcissism encouraged by Filterworld.

Depp 发现，有些顾客在她店里看到她没有把最新的超人气 TikTok 书籍摆放在显眼位置，甚至可能根本没有摆放在店里时，都会感到惊讶。他们期待她的店面能遵循信息流的逻辑，但事实并非如此。这是因为 Depp 亲自挑选店里的所有商品：她用一种算法。“他们没有意识到，策展并非由互联网上的第三方完成，因为他们看到的所有信息都是由互联网上一只看不见的手策划的，”她说道。在充分迎合算法信息流和过度依赖它们之间，界限很难把握。追求极致关注，进而获取利润的诱惑始终存在，但过度依赖自动化信息流的累积效应是一种脱敏。最终，我们无法想象文化除了算法之外还能以其他任何方式运作。

Depp finds that some people coming into her stores are surprised when she doesn’t have the latest hyperpopular TikTok book displayed front and center, or perhaps even in the store at all. They expect her space to follow the logic of the feeds, but it doesn’t. That’s because Depp selects everything in the store herself: an algorithm of one. “They don’t realize the curation isn’t being done by some third party on the Internet, because the way they see all information is something that’s been curated by an invisible hand on the Internet,” she said. The line between catering to algorithmic feeds enough and relying on them too much is a hard one to walk. The temptation to court the extreme attention, and thus profit, is always there, but the cumulative effect of so much reliance on automated feeds is a kind of desensitization. We end up not being able to imagine culture operating any other way than algorithmically.

算法对创造力的压力

ALGORITHMIC PRESSURE ON CREATIVITY

我第一次见到我的朋友哈莉·贝特曼是在2010年代初，当时我们都住在布鲁克林。她从湾区搬来，一边做咖啡师，一边从事绘画和编辑插画的艺术创作。我们是通过一位共同的朋友认识的，这位朋友经营着一个博客，哈莉（用她的姓氏感觉太正式了）在那里投稿漫画，我负责写文章。但实际上，我们每天在推特和Instagram上关注彼此的作品。当她在像The Awl这样的出版物上发表艺术作品和文章时，我为她感到高兴。The Awl是一个颇具影响力的纽约博客，于2018年关闭，后来又在《纽约客》的网站上发表文章。

I first met my friend Hallie Bateman in the early 2010s when we were both living in Brooklyn. She had moved from the Bay Area and was working as a barista while pursuing her artistic practice of drawing and editorial illustration. We met through a mutual friend who ran a blog where Hallie—it feels too formal to use her last name—contributed comics and I wrote posts. But really it was Twitter and Instagram where we followed each other’s work on a daily basis. I cheered her on when she published art and writing at publications like The Awl, an influential New York blog that shut down in 2018, and eventually the website of The New Yorker.

哈莉作品的奇特和个性总是让我印象深刻。2010年代是干净、扁平化设计的时代，光滑柔和的色彩图案看起来像是经过熨烫机熨烫过的。（一个名为“人类”的推特账号（Hallie 的绘画风格很单一，类似于那种平面的、收集通用的初创公司插画风格的样本，这种风格本来是为在社交媒体上推广而准备的。）Hallie 而是坚定地在纸上作画，潦草的、波浪起伏的线条让人想起 Lynda Barry 和 Roz Chast 等漫画家。干净利落从来都不是重点；相反，她似乎更看重的是她能把多少感情融入到每一个涂鸦中，无论是她狗的速写，还是描绘生命本身混乱本质的叙事场景。她最著名的作品之一是她在 Instagram 上发布的一个白色方块，几个小人像在上面行走，每个人都拖着自己原色的线条。他们似乎都不会相交。在图画的中央，Hallie 写道：“我们相遇真是个奇迹。”这是一幅苦乐参半的作品：我们的生活与世界上其他人的重叠真的很少，然而，我们的生活确实有重叠，这一事实值得庆祝。这幅作品非常受欢迎，传播范围远远超出了哈莉的描述，许多人甚至在不知情的情况下就把它纹在了身上。在《过滤世界》中，它被强行去语境化了。

What always struck me was the quirkiness and personality of Hallie’s artwork. The 2010s was the era of clean, flat design, with smooth, pastel-colored graphics that looked like they had been run through an ironing press. (One Twitter account, called Humans of Flat, collected examples of the generic style of start-up illustration that was primed to be promoted across social media.) Hallie instead drew resolutely on paper, with scribbly, wavering lines that were reminiscent of comic artists like Lynda Barry and Roz Chast. Clean perfection was never the point; instead, it seemed to be how much feeling she could pack into every doodle, whether it was a quick sketch of her dog or a narrative scene depicting the chaotic nature of life itself. One of her best-known images is a white square that she posted on Instagram, occupied by a handful of tiny human figures walking across it, each trailing their own primary-colored line. None of them seem like they’re going to intersect. In the center of the drawing, Hallie wrote, “It’s a miracle we ever met.” It’s a bittersweet piece: our lives really overlap with so few other people in the world, and yet the fact that they do at all is worth celebrating. The piece is so popular, and has spread so far beyond Hallie’s account, that many people have gotten it tattooed on their bodies without even knowing its origin. In Filterworld, it was forcibly decontextualized.

我记得那些夜晚，我们慵懒地围坐在哈莉布鲁克林公寓客厅的矮咖啡桌旁；她的室友们有学者、作家，也有艺术家。那里的氛围鼓励每个人都能发挥创造力：你不必自诩为艺术家，也能画画、演奏乐器或写诗。其他人，无论专业水平如何，也没有权威评判你的作品是否缺乏艺术价值。（这与互联网截然相反，在互联网上，每个人都觉得自己有权批评或评论。）哈莉也通过她的网络形象展现了这种精神——这就是为什么当她开始发帖说她再也无法忍受Instagram时，我感到很惊讶，尽管她在Instagram上积累了十多万粉丝，并利用这个平台出售作品的印刷品，最终还向大型出版商出售书籍。

I remember evenings sprawled around the low coffee table in the living room of Hallie’s Brooklyn apartment; her roommates were academics, writers, and other artists. The atmosphere was such that anyone could be creative: You didn’t need to consider yourself an artist to make a drawing, play an instrument, or write a poem. Nor did anyone else, no matter their level of expertise, have the authority to judge whatever you made as artistically unworthy. (The opposite of the Internet, where everyone feels entitled to critique or comment.) Hallie also exuded that spirit through her online presence—which is why I was eventually surprised when she began posting that she couldn’t stand Instagram anymore, though she had gathered over a hundred thousand followers there and used her platform to sell prints of her work and, eventually, books with major publishers.

哈莉的职业生涯和在线艺术实践的轨迹勾勒出2000年代末和2010年代互联网创造力的历史，当时用户自制的小型网站逐渐被大型社交平台所取代，而算法信息流则决定了受众看到的内容。这也展现了这些信息流的规模和自动化程度如何变得疏远那些最初让这些作品吸引观众的创作者。

The arc of Hallie’s career and artistic practice online sketch out a history of creativity on the Internet over the late 2000s and 2010s, as user-made, small-scale websites gave way to massive social platforms where the algorithmic feed dictated what audiences saw. It also shows how the sheer scale and automation of those feeds have become alienating to the creators who made them appealing to audiences in the first place.

我打电话给住在洛杉矶的哈莉，她当时住在那里，在公寓里有一间工作室，上面挂满了画作——现在她不怎么在网上发表这些画作了。在装有镜子的衣柜上方的一面墙上，她用胶带贴了一些纸条，上面用大写字母写着一句话：我相信你。2007 年，在加州大学圣克鲁斯分校就读期间，她开通了 Twitter 账户；因为感觉像一个私人空间，所以她用了一个很无意义的名字“hallithbates”，从此以后这个名字就一直沿用在她所有的社交媒体上。她第一次在网上分享艺术作品的地方是一个 Blogspot 网站。大学毕业后，她开始使用 Instagram，当时她独自一人在巴黎和巴塞罗那旅行：“它源于那种孤独的感觉，”哈莉告诉我。

I called Hallie in Los Angeles, where she lived at the time and kept a studio space in her apartment covered with drawings—drawings that she doesn’t publish online so often anymore. High on one wall above the mirrored closet, she taped pieces of paper with a simple message written on them in tall capital letters: I believe in you. She started her Twitter account in 2007 while attending U.C. Santa Cruz; since it felt like a private and personal space, she used a nonsense version of her name, “hallithbates,” that has since stuck for all of her social media. The first place she shared her art online was a Blogspot site. She got on Instagram only after college, when she was traveling alone in Paris and Barcelona: “It came out of this feeling of being lonely,” Hallie told me.

当一家科技网站的全职插画工作失败后，她搬到了布鲁克林，开始做咖啡师。Twitter 和 Instagram 成了她分享作品、与其他艺术家联系的平台，她可以听到那些她想结识的同行和同事们用不同的笔和纸交换笔记。社交媒体提供了一个艺术社群，而当她在 Twitter 上结识了最初认识的人时，这种社群也渗透到了现实世界。

When a full-time illustration job for a technology website fell through, she moved to Brooklyn and took barista shifts. Twitter and Instagram became a way to share her work and connect with other artists, listening in as the people she wanted to have as peers and colleagues swapped notes on different pens and paper types. Social media provided an artistic community, which bled into the physical world as she met people she first got in touch with on Twitter.

“那时我并没有创作任何作品发布在Instagram上；我只是拍拍我的素描本。一切都非常随意，”她说。获得的点赞在她职业生涯早期给予了鼓励：人们在关注她，并成为她作品的粉丝，尽管她没有任何艺术院校的学历或机构背景。“感觉就像得到了来自宇宙的赞许，让我继续做我正在做的事情。”哈莉搬到纽约时，她的Instagram粉丝有一千人；到2015年，她的粉丝已经达到两万。她开始扫描自己的画作，更多地编辑图片，并像创作出版漫画一样认真对待Instagram帖子。这个平台本身就成了她的目标。

“I was not making things to post on Instagram at that point; I would just take pictures of my sketchbook. It was incredibly casual,” she said. The likes she got provided encouragement early in her professional career: people were watching and were fans of what she was doing, even though she didn’t have any art-school credentials or institutional affiliations. “It felt like I was getting a lot of thumbs-up from the universe to keep doing what I was doing.” When Hallie moved to New York, she had a thousand Instagram followers; by 2015 she had accrued twenty thousand. She began scanning her drawings, editing her images more, and being as thoughtful about Instagram posts as her comics for publications. The platform became an end in itself.

Hallie 还意识到 Instagram 动态会奖励特定的品质。她一直将视觉艺术与写作相结合，但信息清晰的帖子获得了最多的互动。“如果我发布“看起来很漂亮，但并没有得到太多的回应，”她说道。这种影响不仅仅是算法推送的结果；消费者的品味并不总是与艺术家自己的愿景相吻合。但推送的加速和反馈的即时性会增强艺术家的自我意识。当哈莉开始发布她的“方向”系列时，她的 Instagram 账户增长最快，粉丝从三万增加到六万多。在这个系列中，她用彩色建筑纸上的粗体墨水写下了格言式的人生建议——“不要把悲伤误认为是深度”、“表达你对你爱的人的爱”。

Hallie also realized that the Instagram feed rewarded specific qualities. She had always combined visual art and writing, but posts with clear written messages got the most engagement. “If I posted something pretty to look at, it didn’t get as much of a response,” she said. This effect isn’t solely a consequence of the algorithmic feed; consumers have tastes that don’t always mesh with an artist’s own vision. But the acceleration of the feed and the instantaneousness of the feedback begets an intensified self-consciousness on the part of the artist. Hallie saw the biggest growth of her Instagram account, going from thirty thousand followers up to more than sixty thousand, when she began posting her “Directions” series. For that series, she wrote aphoristic lines of life advice—“Do not mistake sadness for depth,” “Articulate what you love about the ones you love”—in bold strokes of ink on colorful construction paper.

“这是一种格式；它易于识别；你可以重复。我会批量绘制，然后一遍遍地发布，”哈莉解释道。这就像一个模因式的流水线流程，非常适合Instagram：鲜艳的色彩和简洁的文字，除了简单的道德信息外，还为她的粉丝动态增添了一丝趣味。粉丝们开始仅仅因为这些作品就关注她的账号。“人们关注你，并且一直期待着这一点，”她说。每幅“方向”系列作品都能获得数千个赞。单从参与度来看，这个系列堪称成功，但哈莉对此却很矛盾。她的风格一直以来都因作品而异。“我开始对它的流行感到不满和困惑。我不得不告诉自己，‘别受此影响，继续创作。’这一切也开始感觉像是对你其他作品的批评，”她说。她其他作品的点赞数都更少，这是否意味着这些作品更糟糕？她应该永远创作“方向”系列吗？哈莉感受到的压力是要让她的其他艺术作品也同样明亮、清晰和简单，就像音乐家为了让歌曲在 TikTok 上取得成功而要在前面加入副歌，或者作家为了让自己的作品火爆到点亮 Twitter 推送而感受到的压力一样。

“It’s a format; it’s identifiable; you can repeat it. I would draw them in big batches and then go through them and post,” Hallie explained. It was a meme-like assembly-line process perfectly suited for Instagram: the bright colors and simple text added a little spice to her followers’ feeds along with simple moral messages. Followers came to rely on her account for those pieces alone. “People follow you and continue to expect that,” she said. Each “Directions” image would get thousands of likes. The series counted as an unmitigated success purely in terms of engagement, but Hallie was ambivalent about it. Her style had always varied from piece to piece. “I was starting to resent and get confused by the popularity of it. I had to tell myself, ‘Don’t be affected by this, just keep going.’ It all starts to feel like a commentary on your other work, too,” she said. Did the fact that all her other drawings got fewer likes mean that those pieces were worse? Should she just keep making her “Directions” forever? The pressure that Hallie felt to make the rest of her artwork similarly bright, clear, and simple is much like the pressure that a musician feels to frontload the hook of a song so it succeeds on TikTok or a writer feels to have a take so hot it lights up the Twitter feed.

与此同时，Instagram 也在发生变化。Hallie 在 2017 年左右感受到了算法推送的显著变化，就像 Patrick Janelle 和其他人一样。“我开始真正感受到心理上的转变，从最初有趣和鼓舞人心变成了有点厌倦和不知所措。Instagram 变得越来越不稳定，”她说。Instagram 的推荐促使她的作品观众不理解其背景；当她的作品或标题涉及政治问题时，她开始收到仇恨评论。

At the same time, Instagram was changing, too. Hallie felt a marked difference in the algorithmic feed around 2017, just as Patrick Janelle and others did. “I started to actually feel the psychological shift from fun and encouraging to being a little tired of it and overwhelmed by it. Instagram was becoming much more volatile,” she said. Instagram recommendations pushed her work to audiences that didn’t understand its context; when her art or captions addressed political issues, she began to receive hateful comments.

因此，哈莉决定退出社交媒体。她仍然创作许多其他艺术作品，但她不会发布。以前，她的创作过程不可避免地受到网络上热门作品的影响，这得益于持续不断的反馈循环。现在，情况不同了。“我已经从一个‘试探气球’式的艺术家转变为‘我正在创作一大堆东西，除了朋友之外我不会告诉任何人’类型的艺术家，”她说。“如果我不是为了这个平台创作艺术，那我就只是为了我自己，或者为了人类。”

Hence Hallie’s decision to move off social media. She still makes plenty of other art, but she doesn’t post it. Before, her creative process was inexorably shaped by what played well online, with the help of the constant feedback loop. Now, it’s different. “I have transitioned from being a test-balloon-style artist to being the ‘I’m working on a ton of stuff that I’m not telling anyone about except my friends’ type of artist,” she said. “If I’m not making art for this platform, I’m making it just for myself, or humanity.”

在“过滤世界”时代，任何想法或思绪都可以立即公开，并接受测试，以吸引受众，这种内在的创作过程，甚至独立思考的过程，都显得弥足珍贵。艺术家作为影响者，并非内省；她存在于事物转瞬即逝的表面，根据反应不断迭代和调整。Hallie 的评论让我感到一阵深深的悲哀：我是否已经无法真正独立思考，或者说，如果没有隐形受众的激励，我是否不愿进行创造性工作？哲学家韩秉哲曾描述过，生活在后互联网社会的人们可能“不再拥有潜意识”。

That kind of internal creative process, or even the process of thinking on one’s own, is something that feels lacking in the Filterworld era, when any idea or thought can be made instantly public and tested for engagement. The artist-as-influencer isn’t introspective; she exists on the ephemeral surface of things, iterating and adapting according to reactions. Hallie’s comments made me feel a kind of personal grief: Have I been left incapable of truly thinking for myself, or unwilling to do that creative work without the motivation of an invisible audience? The philosopher Byung-Chul Han has described how people living in post-Internet society may “no longer have an unconscious.”

如今艺术的创作和传播方式已与过去几个世纪截然不同。那时，安静的工作室、车间或书桌旁，不再有观众的持续诘问。创新并非源于对持续参与指标的适应，而是源于创造性的飞跃，这些飞跃乍一看或许令人震惊。当你过分提前迎合预期，或为了适应特定的变量而重新安排你的想象力时，这种飞跃可能被扭曲或缩短。这对艺术家和消费者来说都不利。“有时，人们直到你把作品展示给他们看，才知道自己想要什么。我内心深处有些抗拒，不愿直接满足人们的需求，”哈莉说道。

The way that art is made and distributed now is not how it was for the past few centuries. There was no constant heckling from an audience in a quiet studio or workshop or at a writing desk. Innovation came not from adaptation to a metric of constant engagement but in creative leaps that might have been shocking at first glance. When you conform to expectations too much in advance or rearrange your imagination to fit a particular set of variables, it can mean that that leap is denatured or cut short. That’s bad for both artists and consumers. “Sometimes people don’t know what they want until you show it to them. There is some part of me that resists just giving people exactly what they ask for,” Hallie said.

我问哈莉，她是否担心不去追逐更多粉丝会影响她的谋生能力。她解释说，不依赖算法反馈和持续增长的粉丝数量实际上让她觉得自己正在为自己的创作实践奠定一个更稳定的基础。潮流和平台总是在变化，但她确信自己知道自己的工作方向：“如果我适应每一种趋势，如果我跳上每一个新平台，试图在那里吸引粉丝，那我就只能建一座又一座沙堡了。如果算法现在让我们失望了，那就意味着它从来就不稳定。它就像一个酒肉朋友。”

I asked Hallie if she was worried about not chasing more followers hurting her ability to make a living. She explained that not relying on algorithmic feedback and the numbers continually going up actually made her feel like she was creating a more stable base for her creative practice. Trends and platforms always change, but she could be sure she knew where her work was going: “If I adapt to every trend, if I hop on every new platform and try to build a following there, I’m going to be building sandcastle after sandcastle. If the algorithm is failing us now, that means it was never stable. It was like a fair-weather friend.”

沙堡的比喻似乎恰如其分。关注、点赞、互动，以及实时推荐，都过于便捷地将我的兴趣孤立地呈现出来——它们都是暂时的，取决于平台本身，而平台本身不可避免地会随着时间的流逝而变化和消亡。2010年代的近期历史，随着Facebook的崛起和随后的衰落，表明没有哪个社交网络大到不能倒闭，或者被选择遵循一套新的社交或技术规则的竞争对手取代。当这种情况发生时，用户只能自生自灭，因为我们的数字生活更多地是由商业考量而非我们自身的最佳利益所决定的。

The sandcastle metaphor felt apt. The followings, likes, and engagement, the real-time recommendations that presented a too-conveniently siloed version of my interests—they were temporary and contingent on the platforms themselves, which inevitably change and fade with time. The recent history of the 2010s, with the rise and then growing irrelevance of Facebook, has shown that no social network is too big to fail or get supplanted by a competitor that chooses to play by a new set of rules, social or technological. When that happens, users are left to fend for themselves, as our digital lives are dictated more by business concerns than our best interests.

莫莉·拉塞尔

MOLLY RUSSELL

目前，我们用户完全受制于算法推荐和信息流。它们如同弗兰肯斯坦式的现象，由人类发明并赋予力量，但其实际作用却远远超出了预期。我们无法控制或影响它们。我们无法完全退出，同时仍然使用已成为现代成人生活不可或缺的数字平台。如同邮局、下水道系统或输电线一样，它们至关重要，然而，与这些公共基础设施不同，它们不受政府监督或监管，也不受选民决定的约束。推荐系统泛滥成灾。或许我们往往会忽视它们在文化领域的反复无常，因为它们影响的内容似乎不如自来水重要。如果Spotify在用户听完一张Metallica的专辑后开始推荐一连串的金属音乐，那么这种算法的单调性就可以被一笑置之，或者被当作一个小故障忽略。这看起来可能并不特别危险，因为最大的风险是无聊。然而，内容的算法加速可能是一个生死攸关的问题。

Right now, we users are at the mercy of algorithmic recommendations and feeds. They are a Frankensteinian phenomenon, invented and given power by humans but far surpassing their expected role. We cannot control or influence them. We cannot wholly opt out while still using the digital platforms that have become necessary parts of modern adult life. Like the post office, the sewer system, or power lines, they are essential, and yet, unlike such public infrastructure, they aren’t subject to government oversight or regulation, or the decisions of voters. Recommender systems run rampant. Perhaps we tend to overlook their capriciousness within the cultural sphere because the material that they influence seems less important than, say, running water. If Spotify begins recommending a litany of metal music after a user listens to a Metallica album, then the algorithmic monotony can be laughed off or ignored as a glitch. It may not seem particularly dangerous, since the biggest risk is boredom. Yet the algorithmic acceleration of content can be a life-or-death problem.

2017年11月，来自伦敦西北部的14岁学生莫莉·拉塞尔自杀身亡。然而，拉塞尔并非她行为的全部责任人。2022年，北伦敦高级验尸官安德鲁·沃克对她的死因进行了不同的描述：“她因抑郁症和网络内容的负面影响而自残身亡。” 在一篇开创性的沃克正式将社交媒体列为潜在致命疾病，是一种医疗隐患。拉塞尔的经历讲述了社交媒体时代年轻人的一个熟悉故事，那时人们的生活线上线下一样多。像任何青少年一样，拉塞尔花了很多时间使用数字平台。在她去世前的六个月里，她在 Instagram 上接触了超过 16,000 条内容。根据政府对她死亡的调查，其中 2100 条（13%）与自杀、自残和抑郁有关。拉塞尔还在 Pinterest 上收集了 469 幅相同主题的图片。社交媒体信息流可以成为友谊和恋爱关系的载体，也可以根据任何兴趣提供内容。然而，推动漫威电影迷或让歌曲走红的社会和技术力量也可能影响或加剧精神疾病。

In November 2017, a fourteen-year-old student from northwest London named Molly Russell died by suicide. Russell wasn’t wholly responsible for her actions, however. In 2022, Andrew Walker, the senior coroner for North London, described her death in a different way: “She died from an act of self-harm while suffering from depression and the negative effects of online content.” In a pioneering moment, Walker formally cited social media as potentially lethal, a medical danger. Russell’s background tells a familiar story of youth in the age of social media, when life is lived as much online as it is off. Like any teenager, Russell spent a lot of her time using digital platforms. In the six months before her death, she was exposed to more than sixteen thousand pieces of content on Instagram. Twenty-one hundred of those, or 13 percent, were related to suicide, self-harm, and depression, according to the government inquiry into her death. Russell had also assembled a Pinterest board of 469 images around the same themes. Social media feeds can play host to friendships and romantic relationships or deliver content based on any interest. Yet the same social and technological forces that propel Marvel movie fandom or make a song go viral can influence or exacerbate mental illness.

算法推荐在拉塞尔收到如此多与精神疾病相关的内容的过程中发挥了重要作用。《连线》杂志报道称，拉塞尔收到一封名为“你可能喜欢的抑郁症 Pin 图”的电子邮件，其中包含一张带血的剃须刀图片，作为推荐添加到她的图板中——将这些负面图像与占据 Pinterest 大量内容的家居装饰灵感同等对待。后来，Facebook 透露，拉塞尔被推荐了 30 个不同的账户，这些账户的名称中“提到了悲伤或令人沮丧的主题”，这为她提供了更多接收可能恶化其精神健康的内容的渠道。在她自己的秘密推特账户上，拉塞尔转发了名为“抑郁症语录”等账户的推文，分享了美化自杀的图片，并联系了谈论抑郁症的有影响力的人物。这些社交网络信息流导致了验尸官所说的“狂欢期”，在此期间，拉塞尔会消费自动发送给她的内容。

Algorithmic recommendations played a major role in delivering Russell so much mental-illness-related content. Wired reported that Russell had received one email titled “Depression Pins you may like” that included a picture of a bloody razor as a suggestion for her board—treating such negative imagery the same as the home-decor inspiration that occupies so much of Pinterest. Later, Facebook revealed that Russell was recommended thirty different accounts that “referred to sad or depressing themes” in their names, providing even more avenues for receiving content that may have worsened her mental health. On her own secret Twitter account, Russell retweeted accounts with names like “depression quotes,” shared images that glorified suicide, and reached out to influencers who spoke about depression. The social-network feeds drove what the coroner called “binge periods,” when Russell would consume content that was automatically delivered to her.

Russell 的死是算法过度扩张造成的人员伤亡的一部分，算法内容更新速度过快，规模过大，无法手动控制。没有哪家杂志的编辑会发布如此大量的抑郁症内容，电视台也不会播放。但算法推送可以即时生成一个按需内容集，提供 Russell 可能觉得最吸引人的内容，即使这对她有害。Russell 的悲剧表明，Filterworld 的问题往往是结构性的变化，根植于数字平台的运作方式。用户改变自身行为的作用有限；我们不能相信这些机制会将我们的福祉置于激发更多互动以推动广告收入之上。用户在数字平台中只能发挥某些类型的自主性。例如，他们可以追求特定的内容主题，但无法改变推荐算法的方程式。除了算法推送之外，我们没有足够的其他选择来浏览互联网，部分原因是互联网现在被少数几家公司主导。

Russell’s death was part of the human toll of algorithmic overreach, when content moves too quickly at too vast a scale to be moderated by hand. No magazine’s editor would have published a flood of such depression content, nor would a television channel broadcast it. But the algorithmic feed could assemble an instant, on-demand collection, delivering what Russell may have found most engaging even though it was harmful for her. The tragedy of Russell’s case demonstrates how the problems of Filterworld are most often structural, baked into the ways that digital platforms function. Users changing their behavior can only go so far; we can’t trust that the mechanisms will ever prioritize our well-being over sparking more engagement that drives advertising revenue. Users can exert only certain kinds of agency within digital platforms. They can pursue a specific theme of content, for example, but can’t alter the equation of the recommendation algorithm. We don’t have enough alternative options to navigate the Internet outside of algorithmic feeds, in part because the Internet is now so dominated by just a few companies.

结构性垄断

STRUCTURAL MONOPOLIZATION

如今，我们的在线体验高度中心化。消费者被圈养在少数几个大型平台——Twitter、Facebook、Instagram、TikTok、YouTube——并在平台的框架内寻找他们想要发现的一切。由于受众如此专注于这些平台，创作者也必须在这些平台上努力，将他们的作品转化为内容，并使其适应平台的信息流。如果他们不这样做，他们将相对隐形。尽管我们在网上看到的很多内容都是“用户生成”的——自由上传，没有门槛或支持——但它仍然必须符合企业预先设定的模式。然而，互联网在其数十年的历史中经历了数次中心化和去中心化的循环，而更早的时代或许能提供更好的用户体验模式。

Our experiences online today are heavily centralized. Consumers are herded into a handful of massive platforms—Twitter, Facebook, Instagram, TikTok, YouTube—and left to seek out whatever it is they want to discover within the platforms’ confines. Because audiences are so concentrated on those platforms, creators must work within them, too, transforming their work into content and adapting it to platforms’ feeds. If they don’t, they will remain relatively invisible. Though so much of the content we see online is “user-generated”—uploaded freely, without either gatekeeping or support—it still has to fit into preestablished molds determined by corporations. Yet the Internet has gone through several cycles of centralization and decentralization over its decades of history, and earlier eras might provide a better model for user experience.

1969 年，美国国防部下属的国防高级研究计划局（DARPA）控制着屈指可数的联网计算机。阿帕网（ARPANET）是第一个广域计算机网络。它可以将其中包含的每台计算机映射到一张美国地图上——它不仅是集中式的，而且只能通过政府和大学访问。逐渐地，它扩展开来。如果说阿帕网就像一个地铁系统，只连接特定的点，那么创建于 1980 年的 Usenet 则更像是一条公路：任何拥有合适设备的人都可以乘坐。Usenet 是最早通过互联网分发和消费内容的方式之一。它是一个数字广告牌系统，最终连接到了阿帕网。任何拥有服务器的人都可以托管一个“新闻组”，人们可以在其中发布文章和帖子。新闻组提供了讨论各种话题的空间，从最近的政治新闻到家庭酿酒的最佳方法，应有尽有。每个新闻组都有自己的主题和规则；没有任何一个新闻组能够控制整个网络。这些群组的管理和节奏完全由参与者决定。

In 1969, there were just a handful of networked computers under the control of the Defense Advanced Research Projects Agency, or DARPA, part of the US Department of Defense. ARPANET was the first wide-area computer network. It was possible to map every computer that it included on a small diagram of the United States—not only was it centralized, but you could access it only through the government and universities. Gradually, it expanded. If ARPANET was like a subway system, only connecting specific points, then Usenet, created in 1980, was more like a roadway: anyone with the right equipment could get on it. Usenet was one of the earliest ways of distributing and consuming content over the Internet. It was a digital billboard system that eventually connected to ARPANET. Anyone with a server could host a “newsgroup,” which people could use to publish articles and posts. The newsgroups provided space for discussions of anything from recent political news to the best methods for home winemaking. Each one had its own themes and rules; no single newsgroup had control over the network. The curation and pace of these groups were entirely set by their participants.

尽管如此，这些在线受众是一群自我选择的超级用户，他们精通这项新兴技术。他们通常受过良好教育，生活富裕。大多数人第一次接触Usenet以及互联网本身，都是在大学期间，这导致了一个奇怪的问题。每年九月，大批新生涌入新闻组，却对规则和礼仪一无所知。这不是他们的错；他们只是以前从未加入过在线社区。但他们往往会制造混乱，发表离题的评论，参与争吵（“口水战”），就像推土机冲进宁静的森林一样扰乱既定的氛围。渐渐地，这些菜鸟——用我们这一代互联网俚语来说——变得势不可挡，让老用户感到厌烦。一些公司也开始通过家庭拨号连接提供访问Usenet的途径。美国在线（AOL）于1993年将Usenet纳入其中，这批菜鸟的突然涌入被称为“永恒的九月”。这句话的言外之意是，错误的用户群体突然在这些曾经的小众群体中占据了主导地位。数字版图，也就是数字文化的早期版本，似乎已经毁灭，尽管它当然仍然存在。

Still, this online audience was a self-selecting group of power users who were fluent in the nascent technology. They were also likely to be educated and affluent. Most people accessed Usenet for the first time, as well as the Internet itself, at universities when they went to college, leading to a peculiar problem. Every September, a flood of new students would make their way into the newsgroups without knowing the rules or etiquette. It wasn’t their fault; they had simply never been in an online community before. But they tended to wreak havoc, publishing off-topic comments, getting into fights (“flame wars”), and disrupting the established tone like a bulldozer rolling into a peaceful forest. Gradually the noobs—as my generation of Internet slang would put it—became overwhelming, annoying the veteran users. Corporations also began to offer access to Usenet over home dial-up connections. America Online (AOL) incorporated Usenet in 1993, and the sudden influx of noobs became known as “Eternal September.” The implication of the phrase was that the wrong kind of user was suddenly dominant in these once-niche groups. The digital landscape, the early version of digital culture, seemed to be ruined, though of course it persisted.

AOL 将互联网简化为一个简单的选项菜单，并在 2000 年左右达到 2300 万用户的峰值时，将其集中起来。它无处不在的光盘，无论你是否已经拥有这项服务，每月都会通过邮件寄送，并在简单的网页界面上加载一系列横幅广告。点击这些横幅广告，用户可以进入网站指定的娱乐、体育、个人理财等主题区域——类似于 Usenet，但更平易近人，用户生成的内容也更少。这些主题频道是我在 20 世纪 90 年代最早的互联网体验的源泉。当然，我花费时间最多的是“儿童专享”版块，里面有安全的（理论上，但实际上并非如此）儿童聊天室，“家庭作业”帮助”以及带有在线高分排行榜的简陋电脑游戏。

AOL simplified the Internet into an easy menu of options, centralizing it for the peak of twenty-three million subscribers that it reached around the year 2000. Its omnipresent compact discs, which arrived in the mail every month regardless of whether you already had the service or not, loaded a series of banners on a simple web interface. Clicking into them allowed the user to enter into designated website areas for entertainment, sports, and personal finance, among other topics—like Usenet but more approachable and less user-generated. These themed channels were home to some of my own earliest Internet experiences in the 1990s. Of course, the one I spent the most time in was the “Kids Only” section, which hosted safe (theoretically, not in practice) chat rooms for kids, “homework help,” and rudimentary computer games with online high-score charts.

我对早期互联网空间的记忆最为清晰，那是在祖父母家，全家出游时，我通过电脑登录他们的电脑。或许是因为父母在他们家时不太关注我；我可以坐在“电脑室”（客厅外的一个杂物接待室）的小桌子旁，坐在一把中世纪风格的旋转椅上，盯着笨重的白色塑料显示器。电脑室里堆满了CD光盘盒，虽然现在已经完全过时了，里面装着电脑游戏和软件。但发现完全可以在网上找到事情做，而且是在AOL的软件里，它整合了Internet Explorer浏览器，这对我来说是一个惊喜。我被这种可能性深深吸引。AOL频道是我接触到的第一个社交媒体版本，也是我接触在线文化的第一个途径。它们让我感觉到其他人也在通过他们自己的屏幕做着、思考着什么。

I remember these early Internet spaces most clearly from when I would log on via the desktop at my grandparents’ house during family trips. Perhaps it was because my parents paid less attention to me when we were there; I could sit at the small desk in the “computer room,” a miscellaneous antechamber off the living room, on a swiveling mid-century chair and stare into the chunky white plastic monitor. There was a tower of CD jewel cases, now an utter anachronism, holding computer games and software, but it was a revelation to discover that you could find things to do entirely online, within AOL’s software, which incorporated the Internet Explorer browser. I was drawn in by that sense of possibility. The AOL channels were my first version of social media and my first access to culture online. They gave me a sense that there were other people doing and thinking about things through their own screens.

但只要有机会，我便会冒险走出那个公司化的空间，去探索一个更加广阔、更加去中心化的互联网。人们自己搭建HTML网站，没有监管，而且通常缺乏专业性。网络就像一个业余爱好者的天地，充斥着各种自制页面，宣扬着某些特定的粉丝群体（比如电视剧《吉尔莫女孩》）或小众爱好（比如建造独木舟），这些页面很容易在早期的谷歌搜索中被偶然发现。你可以使用像1994年推出的Geocities这样的服务，用基本工具搭建和托管网站，但每个Geocities的页面看起来都不一样。它们是GIF动图在凌乱的框架布局中奇特地拼凑在一起，就像是小孩子做的一样。

But as soon as I could, I ventured out of that corporatized space and found a much wider Internet that was more decentralized again. People built their own HTML websites without oversight and often without much professionalism. The web was an amateur zone made up of handmade pages espousing some particular fandom (say, the TV show Gilmore Girls) or niche hobby (building canoes) that were easy to stumble upon with early Google searches. You could use a service like Geocities, which launched in 1994, to build and host a website using basic tools, but no two Geocities pages looked the same. They were quirky collisions of animated GIFs in messy frame layouts, as though a child had made them.

LiveJournal 于 1999 年推出，它提供了一种更直接的日记记录体验，每个用户的独特之处在于他们发布的文字。它是更流行的博客形式的前身，在这项服务中，用户可以拥有自己的个人页面，并根据自己的个性或选择的主题设计页面。论坛架构也允许人们设计自己的公共空间——vBulletin 和 phpBB 等软件工具的名称让我联想到头像和签名横幅。我加入了我痴迷的大型多人在线角色扮演游戏社区的论坛。

LiveJournal, which launched in 1999, offered a more straightforward diary-keeping experience where what set each user apart was the text they published. It was the predecessor to more popular forms of blogging, a service where users could have their own personal pages and design the presentation to match their personality or chosen theme. Forum architectures allowed people to design their own communal spaces, too—the names of those software tools, like vBulletin and phpBB, evoke for me visions of avatars and signature banners. I joined forums for communities of the massive multiplayer online role-playing games that I was obsessed with.

其中一个网站叫做“商人公会”，是一个金绿相间的公告栏，人们在那里讨论《仙境传说》，这是一款占据了我青少年时期大部分时间的韩国大型多人在线角色扮演游戏。在那里，我可以与那些把我视为同龄人的匿名陌生人互动，这给了我一种我人生中其他时期所缺乏的自主感。高中时，我认真听讲，因为有电脑，所以我可以玩《数字吞噬者》和《俄勒冈之路》。英语课上，我不用举手，而是可以讨论如何为骑士角色装备最佳装备。我在网上学到了很多东西；只是那时这些与我生活中的其他任何事情都无关。我参与的其他论坛专注于特定的即兴乐队或分享音乐会录音。这些是我线下无法接触到的领域，因为我与这些社区缺乏实际的联系。

One of those websites was called Merchant Guild, a green-and-gold bulletin board where people discussed Ragnarok Online, a Korean MMORPG that occupied far too much of my early teen years. There I could interact with pseudonymous strangers who treated me as a peer, which gave me a sense of agency that I lacked in the rest of my life, dutifully paying attention in high school classes where computers existed to play Number Munchers and Oregon Trail. Instead of raising my hand in English class, I could debate how best to equip a knight character. I was learning a lot online; it just wasn’t relevant to anything else in my life at that point. Other forums I participated in were focused on particular jam bands or sharing concert recordings. These were niches I had no way to access offline, because I lacked any physical proximity to the communities.

一般来说，一个网站的存在，离不开人们——一人或多人——的辛勤付出，才让它建立并持续运营。我第一次接触网络出版是在2002年高中，当时我还是高中一年级。当时一位名叫帕克的朋友在她自己搭建的网站上开设了一个博客，用来展示自己的艺术作品。在一周又一周地阅读她的帖子后，我请她也为我做一个，然后花了无数的时间修改横幅设计和字体。最终，我设计出了一个让人联想到咖啡馆内部的场景——我记得网站横幅上有一张经过大量PS处理的冒着热气的咖啡杯照片——或者是一个让人联想到安静的图书馆房间，背景是柔和的灰色和棕色。即使在那时，我经常光顾的唯一一家咖啡馆是一家带免下车窗口的独立星巴克，在那里我大多点的是装着滚烫绿茶的高纸杯，那种氛围对我来说也是一种渴望。

Generally, if a website existed, it was because people—one or many—had worked hard to get it there and keep it running. My first brush with publishing online came in high school, which I entered as a freshman in 2002, when a friend named Parker set up a blog on a website that she had built for herself as an art portfolio. After reading her posts week after week, I asked her to make the same thing for me, then spent endless hours tinkering with banner designs and typefaces. I ended up with something that evoked a café interior—I remember a heavily photoshopped photo of a steaming coffee cup in the site’s banner—or a hushed library room, with a background palette of soft grays and browns. Even then, when the only café I frequented was a freestanding Starbucks with a drive-through window, where I mostly ordered tall paper cups of too-hot green tea, that atmosphere felt aspirational to me.

在真正理解“发表”的含义之前，我已经在那个博客上发了几十篇文章，不过大多是抱怨学校或父母。我甚至没有完全意识到其他人也知道这些令人焦虑的长篇大论。在21世纪初，上网仍然是一件非常无聊的事情，很少有人理解它的含义。在学校里，它并没有引发嘲笑，因为其他孩子不知道论坛是什么；而且，当时的数字生活也更安静，因为那时还没有如今人身攻击猖獗的社交媒体公共空间。成群的喷子或机器人无法骚扰单个用户。一切都规模较小，因为它不那么集中。帕克可能是唯一一个读过我帖子的人。尽管如此，这个博客让我明白了拥有一个数字影子自我意味着什么，一个只存在于网络上的生活和个性的版本。当时，这感觉像是一项彻底的创新，一个令人耳目一新的新奇事物。我可以控制我如何在网上展示自己。尽管如此，我并不太清楚屏幕另一边是谁在关注我，部分原因是我们的网络存在与我们的现实生活联系不那么紧密——在网上用全名会显得很奇怪。在社交媒体占领全球之前，它曾经是一个小众市场，一个鲜为人知的爱好。随着互联网与“现实生活”密不可分，那种脱节感慢慢改变并消失了。

I published dozens of posts on that blog before I really knew what “publish” meant, though they were mostly complaints about school or my parents. I didn’t even fully appreciate that other people were privy to these angsty screeds. Being on the Internet in the early aughts was still a nerdy enough pursuit that few people understood what it meant. It didn’t incite teasing at school because other kids didn’t know what a forum was, and digital life was quieter, too, because the public arenas of social media, where personal attacks are now rampant, did not yet exist. Armies of trolls or bots couldn’t harass an individual user. Everything was smaller in scale, because it was less centralized. Parker was likely the only person who ever read my posts. Still, the blog made me understand what it meant to have a digital shadow self, a version of your life and personality that only existed online. At that time, it felt like a radical innovation, a refreshing novelty. I could control how I presented myself online. Still, I was less aware of who was paying attention on the other side of that screen, in part because our Internet presences were less tied to our physical lives—going by your full name online would have appeared bizarre. Before it took over the world, social media was once a niche itself, an obscure hobby. That sense of disconnection slowly changed and dissipated as the Internet became inextricable from “real life.”

到2004年，我拥有了一个MySpace账户，但我的高中朋友只有几个，上面最受欢迎的活动就是重新整理你最喜欢的歌曲和最好的朋友列表。Friendster比MySpace更早出现，而且在比我年纪稍大的人群中也很受欢迎，它也一直存在。Facebook最初的目标是大学生群体。和Usenet一样，进入学校是你融入社会的途径。这个如今已是全球通用的社交网络最初是用来联系同学、发布聚会照片或关系更新的，但它很快就变成了一个真空，吸收了其他类型的数字内容，尤其是在它向非大学生群体开放之后。

By 2004, I had a MySpace account, but only a few of my high school friends did, and the most popular activity to be had on it was rearranging lists of your favorite songs and best friends. Friendster, which predated MySpace and was popular with a crowd slightly older than mine, persisted, too. Facebook started out with the demographic of college students. As with Usenet, getting into school was the way you got on. The now-universal social network was initially used to connect with classmates and post party photos or relationship updates, but it quickly became a vacuum that sucked in other kinds of digital content, particularly as it opened up to non-college populations.

好友的状态更新与群组通知、新闻文章和广告交织在一起，并经过算法过滤。Facebook 率先推出这种内容碰撞，与其说是因为用户需要，不如说是出于自身利益——就像沃尔玛或亚马逊一样，如果 Facebook 能够同时提供所有内容，那么用户使用竞争对手服务的理由就更少了。互联网从手工构建 HTML 网站这种完全专业化的形式，走向了一种“一刀切”的模式。Facebook 集个人博客、论坛、新闻推送和照片分享于一体。显然，马克·扎克伯格不想让它仅仅成为一个网站，而是我们数字生活的全部。互联网已经开始凝聚成一小群巨型平台。

Friends’ status updates became interspersed with group notifications, news articles, and advertisements, filtered by the algorithmic feed. The company pioneered this kind of collision of content, less because users demanded it and more because it served its own interests—like Walmart or Amazon, if Facebook could offer everything at once, then its users would have less reason to use a competing service. From total specialization in the form of hand-built HTML sites, the Internet was moving toward a one-size-fits-all approach. Facebook was a personal blog, a forum, a news feed, and a photo dump all at once. It was clear that Mark Zuckerberg didn’t want it to be just a website, but the entirety of our digital lives. The Internet had begun to congeal into a small set of giant platforms.

Facebook 在 2010 年代初确实面临竞争。Twitter 在 Facebook 推出两年后推出，提供实时、对于那些希望在尽可能小的空间里获取尽可能多的信息的新闻爱好者来说，它按时间顺序提供信息，因此最初的 140 个字符限制颇具吸引力。Tumblr 则提供了更多图片和更私密的体验；就像 LiveJournal 的更新版一样，它倾向于提供私人沉思和深奥知识收藏。我最喜欢的一些 Tumblr 用户精选了一些内容，例如引人入胜的动漫截图或中世纪手稿插图。（色情内容的艺术性程度各不相同，这使得 Tumblr 有别于其他社交网络。）但在 Instagram 出现之前，没有什么真正威胁到 Facebook 在网络社交领域的主导地位。我记得 2011 年加入这款应用时，感觉就像一股清风。Instagram 继承了 Facebook 最初的成功，将自己定位为一个可以查看好友动态的工具。但它不像 Facebook 那样试图同时涵盖如此多的格式，因此规模庞大，内容组合混乱，界面凌乱。 Instagram 的动态消息没有算法，只提供单一类型的内容。它是一款极简主义的工具，只做一件事：在一个让照片看起来更美观的环境中分享智能手机照片。

Facebook did have competition in the early 2010s. Twitter, which launched two years after Facebook’s debut, offered a real-time, chronological feed of information for news addicts who wanted to consume as much information in as small a space as possible, hence the appeal of the original 140-character limit. Tumblr offered a more image-heavy and intimate experience; like an update to LiveJournal, it tended to play host to private musings and collections of esoterica. Some of my favorite Tumblrs curated material like evocative anime screenshots or medieval manuscript illuminations. (Pornography, in varying degrees of artsiness, was also plentiful, setting Tumblr apart from other social networks.) But nothing really threatened Facebook’s grip on online socializing until Instagram. I remember the app feeling like a breath of fresh air when I joined in 2011. Instagram did what Facebook so successfully achieved in the first place in positioning itself as the way you could see what your friends were up to. But it didn’t have the same scale, chaotic combination of content, and messy interface that Facebook had accrued by trying to encompass so many formats at once. Instagram’s feed was non-algorithmic and offered a single type of content. It was a minimalist tool for doing one thing: sharing smartphone photos in an environment that made them look good.

扎克伯格很快意识到规模较小的公司（Instagram 只有 13 名员工，没有任何收入）是一个威胁。2012 年初，他向 Facebook 当时的首席财务官抱怨 Instagram 以及 Path 和 Foursquare 等其他新社交应用，“如果它们发展到一定规模，可能会对我们造成很大的破坏。”因此，他的解决方案是收购这些小公司，并提供创始人无法拒绝的巨额资金——他建议 5 亿美元或 10 亿美元。扎克伯格表示，收购这些应用将“给我们一年或更长的时间来整合它们的动态，然后其他公司才能再次接近它们的规模。”换句话说，Facebook 将收购这些公司，将它们合并到 Facebook 生态系统中，然后复制它们的创新功能以抵消竞争。“新产品不会有太大的吸引力，因为我们已经大规模部署了它们的机制，”他说。

Zuckerberg quickly realized that the smaller company—Instagram had only thirteen employees and no revenue whatsoever—was a threat. In early 2012, he complained to Facebook’s then chief financial officer about Instagram and other new social apps like Path and Foursquare, “If they grow to a large scale, they could be very disruptive to us.” So, his solution was to buy the smaller companies, offering such a large amount of money that the founders couldn’t refuse—he suggested $500 million or $1 billion. Buying the apps would “give us a year or more to integrate their dynamics before anyone can get close to their scale again,” Zuckerberg said. In other words, Facebook would acquire the companies, merge them into the Facebook ecosystem, and then copy the features that made them fresh to neutralize the competition. “New products won’t get much traction since we’ll already have their mechanics deployed at scale,” he said.

该公司执行了这项计划。扎克伯格提出以 10 亿美元收购 Instagram，2012 年 4 月，Instagram 首席执行官兼联合创始人凯文·希斯特罗姆接受了这个提议。他认为自己并没有这真是个艰难的选择。正如希斯特罗姆告诉一位 Instagram 投资者的那样：“我觉得我们永远也逃不过马克的怒火。”Facebook 要么收购这家公司，要么切断这家初创公司对 Facebook 更大的平台（包括其商业软件和社交数据）的访问，从而确保其增长放缓。他们的策略是“要么收购，要么埋葬”。

The company executed that plan. Zuckerberg offered to buy Instagram for $1 billion, and in April 2012, the CEO and cofounder of Instagram, Kevin Systrom, accepted it. He didn’t think he had much of a choice. As Systrom told one Instagram investor, “I don’t think we’ll ever escape the wrath of Mark.” Either Facebook would acquire the company, or it would make sure its growth slowed down by cutting off the start-up’s access to Facebook’s much larger platform, with its business software and social data. The strategy was “buy or bury.”

收购后的几年里，Instagram 基本保持不变。但在 2015 年，这款应用加入了广告功能，2016 年改用算法推送，2017 年又添加了 Stories，这是一种短篇帖子功能，旨在抄袭（并摧毁）其竞争对手 Snapchat。Instagram 逐渐变得越来越像 Facebook：各种内容类型的大杂烩；个人和非个人的社交联系；推送内容远不如公司当时强调的重点，比如视频或购物机会。对于一款致力于打造美学体验的应用来说，这些混乱的变化令人沮丧。2018 年末，希斯特罗姆和 Instagram 联合创始人迈克·克里格彻底离开了 Facebook。（“没有人会因为一切都很棒而辞职，”希斯特罗姆当时说道，暗示了他的不满。）

For several years after the acquisition, Instagram remained more or less the same. But in 2015 the app incorporated advertising, in 2016 it switched to an algorithmic feed, and in 2017 it added Stories, the ephemeral posts that were its attempt to copy (and destroy) its competitor Snapchat. Instagram gradually became more and more like Facebook itself: a mishmash of different content types; personal and impersonal social connections; and a feed that gave you less what you wanted than whatever the company was emphasizing at the moment, like videos or shopping opportunities. The chaotic changes were frustrating for an app that had been devoted to aesthetic experiences. In late 2018, Systrom and his Instagram cofounder Mike Krieger left Facebook entirely. (“No one ever leaves a job because everything’s awesome,” Systrom said at the time, hinting at his dissatisfaction.)

早在Facebook收购Instagram时，我们就感觉到互联网的围墙似乎正在将我们这些用户紧紧地包围起来。像Geocities这样的网络，或是Tumblr这样的个人表达平台，原本广阔的可能性和混乱都被封闭了。数字生活变得越来越模板化，变成了一堆需要填写的方框，而不是一块可以展现自我形象的画布。（你不用重新设计你的Facebook个人资料；你只需更换你的头像即可。）我感到一丝失落，但起初，用创造力换取传播影响力似乎是值得的：你可以在社交媒体上同时与很多人交流！但这种曝光也变得令人精疲力竭，我怀念以前的亲密感，怀念互联网作为一个私人空间——一个逃避现实生活的避难所，而不是现实生活的决定性力量。随着围墙的关闭，算法推送的影响力和权威性也越来越强。

Already, when Facebook bought Instagram, it felt as though the walls of the Internet were closing in a little tighter around us users. The broad expanse of possibility, of messiness, on a network like Geocities or the personal expression of Tumblr was shut down. Digital life became increasingly templated, a set of boxes to fill in rather than a canvas to cover in your own image. (You don’t redesign how your Facebook profile looks; you just change your avatar.) I felt a certain sense of loss, but at first the trade-off of creativity for broadcast reach seemed worthwhile: You could talk to so many people at once on social media! But that exposure became enervating, too, and I missed the previous sense of intimacy, the Internet as a private place—a hideout from real life, rather than the determining force of real life. As the walls closed in, the algorithmic feeds took on more and more influence and authority.

创新放缓。各种涌现的其他工具和平台很快被 Facebook 和谷歌挤占。Twitter 在 2013 年创建了短视频网站 Vine，但由于管理不善，它在 2017 年关闭了。谷歌自己的社交网络 Google Plus 从 2011 年推出到 2019 年悄然消失，一直令人费解。Kickstarter 和 Patreon 等众筹服务允许支持者直接为他们想看的内容付费，为更多小众文化项目带来了希望，但它们并没有像 Snapchat 和 TikTok（Facebook 社交主导地位的唯一真正竞争对手）那样取得后来的那种势头。

Innovation slowed down. The various other tools and platforms that popped up were quickly crowded out by Facebook and Google. Twitter created the short-form video site Vine in 2013, but mismanagement shut it down in 2017. Google’s own social network Google Plus was incomprehensible from its launch in 2011 until it quietly disappeared in 2019. Crowdfunding services, like Kickstarter and Patreon, offered hope for more niche cultural projects by allowing supporters to pay directly for what they wanted to see, but they didn’t achieve the kind of momentum that Snapchat and TikTok—the only real competitors to Facebook’s social dominance—later did.

很快，Facebook 面临的下一个威胁出现了：即时通讯。该公司发现，新的应用正在利用智能手机用户的联系人列表，取代短信和 iMessage 成为人们直接沟通的方式（这些通讯应用使用互联网连接，而非手机信号）。Facebook 拥有自己的通讯工具 Messenger，该应用于 2011 年首次推出，但该公司担心，一款通讯应用可能会扩展到更大规模的社交网络，从而抢走用户。

Soon the next threat to Facebook emerged: instant messaging. The company was seeing new apps take advantage of smartphone owners’ lists of contacts, supplanting SMS and iMessage as the way people communicated directly with each other (the messaging apps used Internet connection rather than phone signal). Facebook had its own messaging tool, an app called Messenger that first launched in 2011, but the company was rightly concerned that a messaging app could branch into larger-scale social networking and thus draw users away.

2013 年初，一位 Facebook 副总裁写道，即时通讯“是转型为 Facebook 最危险的滩头阵地之一”。2009 年成立的即时通讯应用 WhatsApp 证明了他们的观点，其知名度日益提升，尤其是在亚洲和欧洲。到当年年底，WhatsApp 用户超过 4 亿，收入来自下载和订阅费，而非广告。于是扎克伯格又提出了一个无法拒绝的报价。2014 年初，Facebook 同意以 190 亿美元收购 WhatsApp——尽管 WhatsApp 最近的估值为 15 亿美元——对于一家初创公司来说，这个价格高得惊人。结果，WhatsApp 的性质发生了改变。2020 年，WhatsApp 开始涉足广告业务，但并未发展成为与 Facebook 竞争的社交网络，因为 Facebook 掌控着 WhatsApp 的目标和抱负。扎克伯格可以阻止 WhatsApp 扩展功能——此次收购化解了对母公司的真正威胁。

Messaging “is one of the most dangerous beach heads to morph into Facebook,” one Facebook vice president wrote in early 2013. WhatsApp, a messaging app founded in 2009, proved their point, becoming increasingly prominent, particularly in Asia and Europe. By the end of that year, it had over four hundred million users, with revenue coming from download and subscription fees rather than ads. So Zuckerberg made another unrefusable offer. In early 2014, Facebook agreed to acquire WhatsApp for $19 billion—though WhatsApp had most recently been valued at $1.5 billion—an incredibly high price to pay for a start-up company. In turn, WhatsApp was denatured. In 2020, it incorporated advertising, but it never turned into a social network to rival Facebook, because Facebook managed the company’s goals and ambitions. Zuckerberg could stop it from expanding its features—the acquisition neutralized the real threat to the parent company.

Instagram 和 WhatsApp 只是 Facebook 数十项收购中的两项。谷歌也在 2006 年收购了 YouTube，并将这个视频上传网站打造成了媒体消费巨头，取代了有线电视。其他社交网络则未能幸免。例如，曾经与 Twitter 和 Facebook 旗鼓相当的 Tumblr，2013 年被雅虎以 11 亿美元收购。然而，多年来，它管理不善，增长放缓，几乎没有改变其最初的产品。最终，它于 2019 年以 300 万美元的价格再次卖给了 WordPress。在其他情况下，Instagram 这样的应用或许也会遭遇同样的命运。但我们没有机会一探究竟——它被 Facebook 的引力所吸引。Facebook 变得太大了——不仅仅是大到不能倒，而且可能大到非法。许多内部业务细节来自美国联邦贸易委员会在 2020 年提起的一场诉讼，该诉讼称 Facebook 通过其“反竞争行为”在社交网络领域形成了垄断。许多州政府也加入了这场诉讼或提起了类似的诉讼。

Instagram and WhatsApp are just two of Facebook’s dozens of acquisitions. Google similarly acquired YouTube in 2006 and turned the video-uploading site into a media-consumption juggernaut, a replacement for cable television. Other social networks didn’t survive. Tumblr, for example, once on par with Twitter and Facebook, was bought by Yahoo in 2013 for $1.1 billion. Yet it suffered through years of mismanagement and declining growth, barely changing its initial product. It finally sold again in 2019 to WordPress for just $3 million. Perhaps the same fate would have awaited an app like Instagram under different circumstances. But we did not have the chance to find out—it got sucked into Facebook’s gravity. Facebook became too big—not just too big to fail, but perhaps illegally big. Many of these internal business details come from a lawsuit laid out by the US Federal Trade Commission in 2020 arguing that Facebook had become a monopoly in the space of social networking through its “anticompetitive conduct.” Many state governments have joined the lawsuit or mounted similar ones, too.

经历互联网的这些循环，难免会让人感到些许的冲击。我们用户原本以为应该以某种方式行事，但事实却截然相反，就像从假名到真名的转变一样。我们被要求使用工具构建自己的空间，自由表达自我，却又被命令融入社交网络预设的框架。然而，一旦一种标准占据主导地位，它似乎就失去了控制力。数字平台没有目的论的弧线；它们不会像硬盘随着时间的推移能够存储越来越多的数据那样，朝着一个方向不断完善。相反，它是周期性的，像钟摆一样在中心化和去中心化的不同策略之间摇摆。

There’s a certain amount of whiplash that comes with experiencing these cycles of the Internet. We users think we’re supposed to behave one way, and then the opposite becomes true, like the movement from pseudonyms to real names. We’re asked to use tools to build our own spaces, to freely express ourselves, and then commanded to fit within a preset palette determined by a social network. Yet as soon as one standard becomes dominant, it seems to lose its grip. There is no teleological arc for digital platforms; they don’t move in one direction toward perfection, the way hard drives have been able to store more and more data over time. Instead, it is cyclical, swinging between different strategies of centralization and decentralization like a pendulum.

创新始于小规模，一群用户或一款新应用促成了新的行为模式，例如发布新内容或创建新社区。一度，数字文化感觉自然而令人兴奋，与以往截然不同。但随后，这些新的行为和功能被更大的公司采用，通过抄袭、商业压力或并购。用户带着对旧应用的全新兴奋感加入其中，但随着创新被货币化，新鲜感逐渐消退。任何新表达形式带来的乐趣都会被无情地利用，最常见的形式是增加广告投放。整个过程就像西西弗斯的循环。真实的网络文化总是被破坏，用户难免会抱怨，但也总是会再次出现。没有哪个平台是绝对安全的。最大的现有企业可能会受到一个小型新来者的威胁，仅仅是因为轻微的技术进步——比如 Snapchat 的短暂帖子或 TikTok 的完全算法信息流——或者人们不可避免的事实就是感到无聊，而技术就像时尚一样，必须不断变化才能保持对用户注意力的控制。

Innovation starts at a small level, with a group of users or a new app enabling new forms of behavior, like publishing a new kind of content or creating new communities. For a time, digital culture feels organic and exciting, different from what came before it. But then those new behaviors and features are adopted by larger companies, by copying, by business pressure, or by mergers and acquisitions. Users jump on board with newfound excitement about their old apps, but the newness fades as the innovations are monetized to death. Any joy in the new forms of expression is ruthlessly exploited, most often in the form of increased advertising. There’s something Sisyphean in the whole process. Authentic online cultures are always being ruined, as users inevitably complain, but also always emerging again. No platform is ever completely safe. The biggest incumbent could be threatened by a tiny newcomer, simply because of a slight technological evolution—like Snapchat’s ephemeral posts or TikTok’s wholly algorithmic feed—or the unavoidable fact that people simply get bored, and technology, like fashion, must constantly change to maintain its hold over its users’ attention.

然而，当今时代的互联网从未显得如此单一。各个网站已被淹没在层出不穷的信息流中。所有内容都必须适应相同的几个模式。内容创作者或许可以选择自己的平台，但这些平台本身却越来越趋同，运作方式也大同小异。感觉我们的选择已经缩减到了一个令人窒息的地步。互联网巨头们规模庞大、实力雄厚，拥有数十亿用户和数千亿美元的市值，即使是创新型初创企业，也难以在不被某种方式击垮的情况下蓬勃发展。他们的优势过于强大，而且没有任何动力去改变那些正在扁平化我们体验的榨取方式。正如联邦贸易委员会在起诉Facebook的文件中所说：“其结果是竞争减少、投资减少、创新减少，用户和广告商的选择也越来越少。”

Still, the Internet in its current era has never looked more monolithic. Individual websites have been subsumed into ever-flowing feeds. All content has to fit into the same few molds. Content creators may have their choice of platform, but the platforms themselves increasingly resemble one another and function in similar ways. It feels like our options have narrowed to a choking point. The Internet incumbents are so large and powerful, with billions of users and market capitalizations in the hundreds of billions of dollars, that even innovative start-ups have a hard time thriving without being crushed in one way or another. Their advantage is too strong, and they don’t have any incentive to change the extractive methods that are flattening our experiences. As the FTC’s lawsuit against Facebook argued, “The result is less competition, less investment, less innovation, and fewer choices for users and advertisers.”

我们用户无法独自对抗这种令人窒息的环境。在应用之间切换和切换设置所能达到的效果有限。要打破“过滤世界”，变革必须在行业层面，在科技公司自身的规模上进行。去中心化往往会赋予用户最大的自主权，尽管它也会给个人带来更高的劳动负担和责任。这也是抵制“过滤世界”、培育数字生活新可能性的最佳方式。但企业不太可能主动接受去中心化，因为通常利润更低。唯一的变革途径或许是强制他们。

We users can’t fight against this stultifying environment on our own. Switching between apps and toggling settings can accomplish only so much. To break down Filterworld, change has to happen on the industrial level, at the scale of the tech companies themselves. Decentralization tends to give users the most agency, though it also places a higher burden of labor and responsibility on the individual. It’s also the best way to resist Filterworld and cultivate new possibilities for digital life. But companies are unlikely to embrace decentralization on their own, because it’s usually less profitable. The only path for change may be to force them.

对于如此庞大而强大的行业，影响着数十亿人，社交网络业务却很少受到政府监管。它似乎陷入了硬件行业和传统媒体行业的鸿沟：硬件行业设备和制造供应链都面临严格审查；而传统媒体行业自美国宪法赋予言论自由权以来，企业可以播放哪些类型的内容就一直是法律问题。社交网络是否应该像报纸和电视台一样，对所托管的所有内容负责？在他们的领域内？他们早就逃避了这一责任。或者，他们应该更像电话线，理论上是中立的信息传输者？但考虑到他们的算法判断，他们显然并非中立。又或者，社交媒体或许属于不良行业，受到严格监管，旨在保障可能滥用它的个人安全。毕竟，这么多用户都上瘾了。

For such a vast and powerful industry, affecting billions of people, the business of social networks doesn’t face much government regulation. It seems to fall into a gulf between the hardware industry, where devices and manufacturing supply chains face scrutiny, and the traditional media industry, where what kinds of content businesses can broadcast has been a legal issue since the US Constitution enshrined free speech. Should social networks be treated like newspapers and television channels, responsible for everything hosted within their domains? They have long escaped that responsibility. Or should they be classified more like telephone lines, theoretically neutral transmitters of information? But they are decidedly not neutral, given their algorithmic judgments. Or perhaps social media belongs in the category of vice industries, with tightly regulated limits meant for the safety of individuals who might otherwise abuse it. After all, so many users are addicted.

无论我们如何对构成 Filterworld 的数字平台进行分类，它们显然都需要某种形式的监管。作为用户，我们只能感受到这些结构带来的后果，并根据这些后果调整自己的行为。新的行为方式，以及由此产生的新的文化形式，需要新的结构，而这些结构只有在科技公司的垄断和双头垄断被打破后才会出现。

No matter how we classify the digital platforms that make up Filterworld, it’s clear that they need some form of regulation. As users, we only feel the consequences of their structures and adapt our behavior to them. New kinds of behavior, and thus new forms of culture, require new structures, which won’t exist until the tech companies’ monopolies and duopolies are disrupted.

寻求透明度

SEEKING TRANSPARENCY

改变数字平台运作方式的最快方法或许是强制透明度：迫使公司解释其算法推荐如何以及何时生效。透明度至少能让用户更多地了解算法如何持续做出推荐内容的决定。如果我们了解算法的工作原理，或许就能更好地抵制它们的影响，做出自主的决定。

The quickest way to change how digital platforms work may be to mandate transparency: forcing the companies to explain how and when their algorithmic recommendations are working. Transparency would at least give users more information about the decisions constantly being made about what to show us. And if we know how algorithms work, perhaps we’ll be better able to resist their influence and make our own decisions.

2016年唐纳德·特朗普当选后，美国公众开始稍微意识到我们是如何被算法信息流操纵的。民主党人无法理解为什么有人会投票给特朗普，因为他们的Facebook和Twitter信息流并没有那么多来自政治光谱另一端的帖子，这形成了伊莱·帕里瑟的“过滤泡沫”——数字回音室。在网上，他们沉浸在一种完全一致认为特朗普荒谬的幻觉中。与此同时，他的支持者周围充斥着强化他们自身观点的内容——另一种形式的同质化。推荐系统将受众划分为两类，两类之间无需重叠，而在人工编辑的报纸或电视新闻节目中，相互曝光的可能性可能更大。这种可能性更大，但并非必然：传统媒体也可能偏向同质化。例如《纽约时报》不愿报道特朗普获胜的可能性。

After the 2016 election of Donald Trump, the American public became slightly more aware of how we were being manipulated by algorithmic feeds. Democrats couldn’t understand how anyone had voted for Trump, given that their Facebook and Twitter feeds didn’t promote as many posts from the other side of the political spectrum, creating one of Eli Pariser’s filter bubbles, a digital echo chamber. Online, they lived in an illusion of total agreement that Trump was ridiculous. At the same time, his supporters were surrounded by content that reinforced their own views—another form of homogeneity. Recommender systems had sorted the audiences into two neat categories that didn’t need to overlap, whereas in a human-edited newspaper or television news program, some mutual exposure might have been more likely. More likely but not guaranteed: traditional media can be biased into homogeneity just as well, as in The New York Times’s unwillingness to cover the possibility of a Trump win.

我在本书前面提到过对过滤气泡的批评；这种现象或许比特朗普胜选本身更令人意外。自由派人士只是因为社交媒体而没有意识到他的受欢迎程度，没有足够重视他作为威胁，从而让他更容易获胜。

I recounted the criticism of filter bubbles earlier in this book; the phenomenon may have done more to cause the surprise of Trump’s win than the fact that it happened. Liberals were simply caught unaware of his popularity in part due to social media, failed to take him seriously enough as a threat, and thus made it easier for him to win.

特朗普确实利用了算法技术。他的竞选团队有效地利用了Facebook的定向广告项目，向那些在线行为表明可能被他的政治观点说服的选民推送信息。他的竞选团队在11月之前的五个月里花费了4400万美元，购买了590万个Facebook广告位，比希拉里·克林顿竞选团队（购买了6.6万个广告位）多出许多倍。特朗普的团队与Facebook紧密合作，使用定向广告软件测试哪些信息效果最好。Facebook广告通常是根据结果而不是显示的次数来购买的；客户支付点击率和政治捐款等行动的转化率。特朗普竞选团队几乎可以肯定，算法推送的信息会对他们有利。

Trump did take advantage of algorithmic technology. His campaign used Facebook’s targeted advertising program to great effect, pushing messages to voters whose online actions showed that they might be convinced by his politics. His campaign bought space for 5.9 million Facebook ads, spending $44 million in the five months before November, many times more than Hillary Clinton’s campaign, which bought 66,000 ads. Trump’s team worked closely with Facebook itself, using the targeted advertising software to test which messages had the most effect. Facebook ads are often bought based on outcomes rather than how many times they are displayed; the client pays for click-throughs and conversions to actions like political donations. The Trump campaign was all but guaranteed that the algorithmic feed would work in their favor.

从用户角度来看，根据用户关注的内容或关注对象，区分哪些 Facebook 内容是付费广告、哪些内容更接近自然内容也变得更加困难。信息流变得更加混乱，因此也更加令人困惑，也更具操控性。这些因素共同导致了公众对 Facebook 的罕见反弹。

On the users’ side, it had also become more difficult to tell which Facebook content was promoted as a paid advertising and which was more organic, based on who or what you follow. The feed had become more chaotic, and thus more confusing and more manipulative. The combination led to a then-rare public backlash against Facebook.

2016年11月，一位名叫克里希纳·加德（Krishna Gade）的工程师被聘用，他曾在Pinterest和Twitter担任管理职位。他成为Facebook负责新闻推送的工程经理，负责其内容排名系统。这在当时至关重要。“关于新闻推送的运作方式，人们提出了更多内部问题，”加德告诉我。他意识到了新闻推送的同质化力量：“有了推荐算法，你会随着时间的推移得到相同的东西。我们如何打破这些模式？”

An engineer named Krishna Gade, who had previously held management positions at Pinterest and Twitter, was hired in the midst of this drama, in November 2016. He became Facebook’s engineering manager for the news feed, working on its content ranking system. It was a vital job at that moment. “There were a lot more inner questions being asked around how the news feed works,” Gade told me. He recognized the homogenizing force of the feed: “With recommendation algorithms, you get the same kind of things over time. How do we break those patterns?”

首先，Gade 开发了一个内部调试工具来了解推荐算法究竟是如何运作的？它可以解释为什么某条内容在特定时间被推荐。每条内容都会附加一个小链接，展示其出现在信息流中的几个原因，以及让算法记录下来的变量。早期产品的截图展示了一些解释，例如用户与发帖人是好友，用户倾向于更频繁地评论带照片的帖子，或者某个帖子在用户所属的特定群组中很受欢迎。其逻辑很简单，主要围绕着参与度这个主要指标——已经很受欢迎的内容会获得更多曝光。但这项功能至少让信息流看起来更连贯。前提是，你得费心点击按钮并进行调查。

First, Gade developed an internal debugging tool to understand how exactly the recommendation algorithm worked. It could tell why a piece of content was being promoted at a given time. Pieces of content came appended with a small link that displayed a few reasons for its appearance in the feed, the variables that made the algorithm register it. Screenshots from the early product show explanations like the user being friends with the poster, the user tending to comment more often on posts with photos, or a post being popular within a particular group that the user is a member of. The logic was basic, mostly revolving around that dominant metric of engagement—what is already popular gets even more exposure. But the feature at least made the feed seem more coherent. That is, if you bothered to hit the button and investigate.

对加德来说，这项功能是数字平台运作方式的基石。“用户应该有权询问正在发生的事情，”他告诉我。这一原则被称为“算法透明度”，认为我们与之交互的算法中的变量和权重应该公开透明，就像观察钟面上驱动齿轮一样。正如迈克·安纳尼和凯特·克劳福德在2016年发表于《新媒体与社会》杂志的一篇论文中所写，理想的透明度是“一种及时观察、理解和管理复杂系统的方法”。了解某些内容被推荐的方式和原因，或许有助于消除我们在线体验中弥漫的算法焦虑，因为我们可以确定推荐考虑了我们的哪些行为。然而，这仍然需要用户自己负责：了解算法的工作原理并不等同于能够控制它。“单靠透明度无法创建负责任的系统，”安纳尼和克劳福德写道。

For Gade, the feature was fundamental to how a digital platform should work. “Users should be given the right to ask for what’s going on,” he told me. This principle goes by the label of “algorithmic transparency,” which argues that the variables and weights that go into the algorithms we interact with should be made accessible and public, like peering at the gears driving a clock face. In its ideal form, transparency is “a method to see, understand and govern complex systems in timely fashions,” as Mike Ananny and Kate Crawford wrote in a 2016 paper in the journal New Media & Society. Knowing how and why something has been recommended might help to dispel the air of algorithmic anxiety that surrounds our online experiences, since we could identify which of our actions the recommendations are considering. Still, it leaves users responsible for themselves: Knowing how the algorithm works isn’t the same as being able to control it. “Transparency alone cannot create accountable systems,” Ananny and Crawford wrote.

2015年，美国联邦贸易委员会（FTC）成立了技术研究与调查办公室，其职责包括研究算法透明度的可能性。即使在那时，算法在人们生活中扮演着越来越重要的角色，这一点显而易见：“消费者每天都在与算法互动，无论他们是否意识到这一点。迄今为止，我们对这些算法的运作方式、背后的动机、数据的使用方式以及数据的结构知之甚少，”时任FTC首席技术专家的Ashkan Soltani在2015年接受《PC World》采访时表示。但迄今为止，相关研究进展甚微。在这方面，尤其是在社交媒体领域。联邦贸易委员会的行动更多地侧重于监管数字广告、加密货币和隐私，而不是信息流的运作方式。

In 2015, the FTC established the Office of Technology Research and Investigation, whose mandate included looking into the possibilities of algorithmic transparency. Even at that point, it was clear that algorithms were playing an increasing role in people’s lives: “Consumers interact with algorithms on a daily basis, whether they know it or not. To date, we have very little insight as to how these algorithms operate, what incentives are behind them, what data is used and how it’s structured,” Ashkan Soltani, then the FTC’s chief technologist, told PC World in 2015. But little progress has been made in that direction, particularly in the context of social media. The FTC office’s actions have focused more often on regulating digital advertising, cryptocurrency, and privacy than on how feeds work.

自几年前推出解释功能以来，Facebook 在透明度方面进展不大。我花了几个月时间在自己的 Facebook 动态中试用该功能，它给出的最常见解释是，某篇帖子“与你看过的其他帖子相比很受欢迎”。几乎任何平台上任何算法推荐的内容都可以做出这样的假设。

After implementing the explanation feature years ago, Facebook has not gone far in embracing transparency. As I’ve experimented with it over months of my own Facebook feed, the most common reasoning it gave me was that a post is “popular compared to other posts you’ve seen.” One could make that assumption about pretty much any algorithmically recommended piece of content found on any platform.

就在2021年，Facebook母公司Meta的全球事务总裁尼克·克莱格（他曾任英国副首相）仍在规划一条更加透明的道路。“你应该能够与算法对话，并有意识地调整或忽略它做出的预测，”他在Medium上发表的一篇题​​为《你与算法：探戈需要两个人》的文章中写道。克莱格认为，应该能够“通过平台设计中内置的喘息空间，在冷静的状态下改变你的个人算法”。这是一个令人回味的描述，暗示了空间与推荐之间的分离，但他的雇主几乎没有付诸行动。Facebook目前赋予用户对其信息流的少数控制权之一是选择优先添加为“收藏”的账户和群组，这意味着算法会更频繁地推广它们，或者通过“暂停”降低优先级（减少对某个账户的访问，但只是暂时的）。很快，算法就会重新发挥作用。用户无法平衡各种话题，无法选择多看朋友圈而不是新闻，也无法减少对政治的愤怒，转而关注积极的内容。尽管每个人的信息流看起来各不相同，以适应我们的兴趣和习惯，但它们都以公司规定的基本方式运作。

As recently as 2021, Nick Clegg, the president of global affairs at Facebook’s parent company Meta (he was previously the deputy prime minister of the United Kingdom), was still projecting a path toward more transparency. “You should be able to talk back to the algorithm and consciously adjust or ignore the predictions it makes,” he wrote in a post on Medium titled “You and the Algorithm: It Takes Two to Tango.” According to Clegg, it should be possible to “alter your personal algorithm in the cold light of day, through breathing spaces built into the design of the platform.” It’s an evocative description, the suggestion of space separated from recommendations, but his employer has put little of it into action. One of the few controls Facebook currently gives users over their feeds is selecting accounts and groups to prioritize as “Favorites,” meaning that the algorithm will promote them more often, or deprioritize with “Snooze” (seeing less of a particular account, though only temporarily). Soon enough, the algorithm reasserts itself. The user can’t balance a range of topics, choose to see more of her friends than news stories, or dial down the political outrage in favor of positivity. Though each person’s feed looks different, adapted to our interests and habits, they all work in the same fundamental way, dictated by the corporation.

盖德于2018年10月离开Facebook。他决定独自研究算法透明度问题，并创办了一家名为Fiddler的新公司。Fiddler帮助客户分解机器学习模型并洞察其内部。这可能意味着评估算法银行流程的运作方式。为什么Facebook会自动向一位客户提供贷款，而拒绝另一位客户，或者为什么像亚马逊Alexa这样的语音控制应用程序总是误解某个词。如果Facebook允许Fiddler访问其数据和模型，它还可以让我们更深入地了解Facebook推荐算法为何会在信息流中推荐特定内容。

Gade left Facebook in October of 2018. He decided to work on the problem of algorithmic transparency on his own, with a new company that he cofounded called Fiddler. Fiddler helps its clients break down machine-learning models and see inside of them. That could mean evaluating why an algorithmic banking process automatically offers a loan to one customer while denying another, or why a voice-controlled app like Amazon’s Alexa consistently misinterprets a particular word. It could also give a much deeper view of exactly why the Facebook recommendation algorithm promotes a certain piece of content within the feed—that is, if Facebook allowed Fiddler to access its data and models.

Fiddler 创建了一个包含变量和结果的仪表盘，允许其客户进行实验性调整，然后根据结果更新其内部模型。它的软件还可以洞察这些常被描述为黑匣子的算法系统。Gade告诉我，“黑匣子”这个词“有点夸张”。虽然机器学习会开发出人类无法理解的抽象模式，但其他机器却可以感知并表示这些模式。“Fiddler 为你提供了一个窥视其结构的视角，”他说。

Fiddler creates a dashboard of variables and outcomes, allowing its clients to experimentally tweak them and then update their internal models with the results. Its software can also look inside these algorithmic systems that are so often described as black boxes. The term black box “is a bit of an exaggeration,” Gade told me. While machine learning develops abstract patterns that can be incomprehensible to a human being, other machines can perceive and represent them. “Fiddler gives you a lens into the structure,” he said.

Gade 通过视频聊天向我演示了 Fiddler 的仪表盘。公司上传模型和数据后，该软件可以“衡量公平性”，使用具体示例（例如不同性别或种族的用户）进行测试，并确保结果一致。许多推荐算法基于对内容的解读，评估帖子、标题或用户评论等文本，以衡量特定内容的相关性。但算法可能会误解语言。Gade 向我展示了一个案例，其中一个模型将“同性恋”一词赋予了非常负面的含义，这意味着包含该词的内容没有得到优先处理。如果这个词的本意是褒义的，那可能就是一个彻头彻尾的错误——或者，它应该被解读为中性。如果自动内容审核或推荐系统误解了某个词或行为，“你可能会让整个群体都沉默，”Gade 说。这就是为什么能够了解围绕给定算法决策发生的事情如此重要。

Gade gave me a demonstration of Fiddler’s dashboards over video chat. After a company uploads its models and data, the software can “measure fairness,” testing it using specific examples like different genders or races of user and ensuring the outcomes are consistent. Many recommendation algorithms are based on the interpretation of content, evaluating texts like posts, captions, or user reviews in order to gauge the relevance of a particular item. But algorithms can misinterpret language. Gade showed me a case in which a model was assigning the word gay a very negative connotation, meaning that content that included it wasn’t getting prioritized. That could be a complete mistake if the word is meant positively—or perhaps it should be interpreted as neutral. If automated content moderation or recommender systems misinterpret a word or behavior, “You could potentially silence an entire demographic,” Gade said. That’s why being able to see what’s happening around a given algorithmic decision is so important.

Twitter 就不透明的危害提供了实实在在的教训。埃隆·马斯克 (Elon Musk) 于 2022 年收购这家社交网络时，人们希望这位企业家能够帮助这家行动迟缓的服务充分发挥其潜力。然而，随之而来的是一系列随机且不成熟的修改，这些修改往往会恶化用户体验，尤其是在信息流方面。Twitter 的每个版本——无论是在移动应用、网站还是软件上——都会Tweetdeck 的功能似乎有所不同。用户可以切换到“最新”推送，这提示推文会按时间顺序排列，但却发现一些奇怪的异常，比如推文会聚集在一起，只来自付费 Twitter 认证的账户，而且会不断被推荐推文打断。我开始觉得我无法相信算法推送的日常运作，更不用说多年来的运作了。虽然我像依赖实用程序一样依赖这项服务，但它提醒我，我从来都不知道幕后发生了什么。它缺乏透明度，更不用说稳定性了，因为 Twitter 对用户没有真正的责任。

Twitter has provided an object lesson in the damage of nontransparency. When Elon Musk acquired the social network in 2022, the hope was that the entrepreneur would help the slow-moving service fulfill its potential. What ensued was a rash of random and half-baked changes that most often altered the user experience for the worse, particularly when it came to the feed. Each version of the Twitter feed—whether on mobile app, website, or the software Tweetdeck—seemed to function differently. A user could toggle to the “Latest” feed, which suggested that the tweets would be in chronological order, only to find strange aberrations like clusters of tweets only from accounts that had paid for Twitter verification and constant interruption by recommended tweets. I began to feel that I couldn’t trust how the algorithmic feed functioned day to day, let alone over years. Though I had relied on the service like a utility, it was a reminder that I never knew what was happening behind the scenes. There was no transparency, let alone stability, because Twitter had no real responsibility to its users.

出版与推广

PUBLISHING VERSUS PROMOTING

正如数字平台无需解释其算法推送的内容一样，它们也无需对推送内容的推广负责——它们将自身与推荐系统的结果区分开来。它们之所以能够做到这一点，是因为美国1996年颁布的《电信法》，该法包含一项名为第230条的《通信规范法》。第230条推动了互联网在过去几十年的迅猛发展。没有它，我们的数字世界将大不相同。但在社交媒体时代，它也让那些取代传统媒体业务的科技公司得以在缺乏传统媒体保障的情况下运营。第230条将像Facebook这样的开放平台与用户在其上发布的内容区分开来。法律规定：“任何交互式计算机服务的提供者或用户均不得被视为其他信息内容提供者所提供信息的发布者或传播者。”这是一个重要的区别，因为“发布者”对其发布的内容负有法律责任：例如，如果一本杂志刊登了诽谤某人的内容，它可能会遭到其内容提供者的起诉。相比之下，Facebook 可以避免这种风险，因为从技术上讲，发布内容的是用户，而不是平台。

Just as digital platforms aren’t responsible for explaining their algorithmic feeds, they also don’t take responsibility for what the feeds promote—they separate themselves from the outcomes of their recommender systems. They can do that because of the United States’s 1996 Telecommunications Act, which included a Communications Decency Act with a piece called Section 230. Section 230 allowed the Internet to grow exponentially over the past decades. Our digital world wouldn’t be the same without it. But in the social media era, it has also allowed the tech companies that have supplanted traditional media businesses to operate without the safeguards of traditional media. Section 230 makes a distinction between an open platform, like Facebook, and what users publish on it. “No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider,” the law states. This is an important distinction, because “publishers” are legally liable for the content they publish: if a magazine prints something defamatory about someone, for example, it can be sued by its subject. Facebook, in contrast, can wave off the risk because it is the user, not the platform, who is technically publishing the content.

第230条款的先例源于20世纪90年代的两起案件。1991年，Cubby诉CompuServe案始于两家数字媒体公司之间的纠纷。罗伯特·布兰查德（Robert Blanchard）旗下的Cubby公司（一家名为Skuttlebutt的新闻服务出版商）起诉了唐·菲茨帕特里克（Don Fitzpatrick）的出版物。Rumorville 的罪名是发表一篇诽谤 Skuttlebutt 的文章。Rumorville可以在CompuServe主办的在线论坛上找到。CompuServe 是早期的家庭互联网提供商，在 20 世纪 90 年代占据主导地位。Blanchard 起诉了 CompuServe 本身以及 Fitzpatrick。美国地方法院的结论是，CompuServe 只是涉嫌诽谤内容的分销商，而非出版商。换句话说，作为托管方，CompuServe 的行为更像是报摊或书店。它无法控制出版内容，只能控制材料如何到达消费者手中。20 世纪 50 年代的一个案例已经裁定，书店不可能合法审查上架的每一本书，因此无需对书籍内容负责。

The precedent for Section 230 came from two 1990s cases. In 1991, Cubby v. CompuServe began as a conflict between two digital media companies. Robert Blanchard’s Cubby Inc., publisher of a news service called Skuttlebutt, sued Don Fitzpatrick’s publication Rumorville because it had published an article that defamed Skuttlebutt. Rumorville was available on an online forum hosted by CompuServe, an early provider of home Internet that was a major force in the 1990s. Blanchard sued CompuServe itself as well as Fitzpatrick. The US District Court concluded that CompuServe was only a distributor, not a publisher of the allegedly defamatory content. In other words, as a host, CompuServe acted more like a newsstand or bookstore. It didn’t control the contents of what got published, just how the material reached consumers. A 1950s case already ruled that bookstores couldn’t possibly legally vet every book that passed through their shelves and were thus freed from responsibility for what the books contained.

另一起关于数字发行的案件结果有所不同。1995年，纽约最高法院审理了Stratton Oakmont诉Prodigy案。Prodigy服务公司是一家在线论坛的主办方，其中包括一个财务讨论空间。Prodigy的一名用户在论坛上发表了诽谤性评论，针对经纪公司Stratton Oakmont, Inc.及其总裁Daniel Porush。此案与Cubby诉CompuServe案类似，但Prodigy对其服务上发布的实际内容的参与程度更高。该论坛对发布内容有规定，并实施了审核系统，包括自动过滤和人工审核。法院裁定，在这种情况下，Prodigy是出版商，而非分销商，因此对其网站上的内容负有法律责任。“Prodigy显然在对内容做出决定……而此类决定构成了编辑控制，”法院判决书中写道。 Prodigy 依据Cubby v. CompuServe的先例请求撤销此案，但无济于事。

A second case on digital distribution had a different outcome. In 1995’s Stratton Oakmont v. Prodigy, which was heard by the New York Supreme Court, Prodigy Services Company was the host of an online forum, including one space for discussions of finances. One of Prodigy’s users published defamatory comments on the forum, targeting the brokerage firm Stratton Oakmont, Inc., and its president Daniel Porush. This case would seem similar to Cubby v. CompuServe, but Prodigy was more involved in the actual content that was published on its service. The forum had rules about what could be posted and imposed moderation systems, including automatic filtering and human moderators. The court determined that in this context, Prodigy was a publisher, not a distributor, and so was legally responsible for the material on its site. “Prodigy is clearly making decisions as to content…and such decisions constitute editorial control,” the court decision read. Prodigy requested a dismissal of the case, based on the precedent of Cubby v. CompuServe, but to no avail.

这些相互矛盾的案件揭示了一个根本性的悖论：不过滤用户内容的互联网服务受到法律保护，而那些试图过滤内容的服务，即使只是为了基本的质量或安全，也不受保护。互联网公司试图影响内容本身的风险更大。两位国会议员克里斯托弗·考克斯和罗恩·怀登认为这个问题需要解决。“如果这条规则生效，那么互联网就会变成狂野西部，“没有人会有任何动力去维护互联网的文明，”考克斯后来告诉《连线》杂志。考克斯和怀登的第230条允许数字平台发布某些内容，特别是任何“淫秽、猥亵、淫荡、污秽、过度暴力、骚扰或其他令人反感的内容”，而无需为其发布承担责任。换句话说，只要内容看起来是为了用户的普遍利益，在线发布者就可以“善意地”干涉内容。第230条于1996年初由比尔·克林顿总统签署成为法律。

The conflicting cases brought up a fundamental paradox: Internet services that did nothing to filter the content going to users were legally protected, while services that did try to filter the content, even just for basic quality or safety, were not protected. It was riskier for Internet companies to try to influence content at all. Two congressional representatives, Christopher Cox and Ron Wyden, decided that the issue needed to be resolved. “If that rule was going to take hold, then the internet would become the Wild West and nobody would have any incentive to keep the internet civil,” Cox later told Wired. Cox and Wyden’s Section 230 allowed digital platforms to mediate some content, particularly anything “obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable,” without being held liable for its publication. In other words, the online distributors could “in good faith” interfere with content as long as it seemed to be for the general good of the users. Section 230 was signed into law by President Bill Clinton in early 1996.

然而，2010年代的主流社交媒体与第230条的关系截然不同。1996年，互联网对于当时约1600万用户来说仍然是一种相对小众的体验。它在媒体传播中扮演的角色要小得多。如今，Twitter和Facebook已经发展壮大，涵盖了数亿人消费各种媒体的方式，从娱乐到新闻。

Yet mainstream social media in the 2010s bears a very different relationship to Section 230. In 1996, the Internet was still a relatively niche experience for its sixteen million or so users. It played a much smaller role in media distribution. Today, Twitter and Facebook have grown to encompass how hundreds of millions of people consume all forms of media, from entertainment to news.

社交网络通过吸收广告收入、将内容孤立到算法推送中以及调解出版商与其消费者之间的关系，取代了传统出版商。其后果是，传统媒体公司遭受重创，与几十年前相比损失了大部分收入，而出版物现在被迫以更少的员工数量维持其作为负责任的出版商的身份。即使在有限的条件下，传统媒体公司仍然对其发布的每篇内容承担责任。与此同时，数字平台可以以第230条为借口，声称自己根本不是媒体公司。

Social networks displaced traditional publishers by absorbing advertising revenue; siloing content into algorithmic feeds; and mediating the relationship between publishers and their consumers. The consequence is that traditional media companies have been decimated, losing most of their revenue when compared to decades past, and publications are now forced to maintain themselves as responsible publishers with much smaller numbers of staff. Even with their restricted circumstances, traditional media companies continue to hold responsibility for every piece of content they publish. Meanwhile, digital platforms could claim they were not media companies at all with the excuse of Section 230.

角色之间的界限变得模糊。在这个新的生态系统中，数字平台承担了出版商的部分职能，决定哪些内容能够触达消费者。如果说CompuServe可以声称自己在诉讼中保持中立，因为它没有影响Rumorville发布的内容，那么Facebook就远没有那么中立了。该算法的策展行为类似于报纸选择将哪些内容放在头版。第230条充当了一道屏障，使社交网络与个人用户在其平台上发布的内容保持距离。这些内容涵盖了从#MeToo调查到种族主义言论或暴力威胁等各种内容。这项法律已经成为社交网络正受到越来越多的质疑，诉讼也试图要求社交网络对其传播的内容负责。

The separations between roles blurred. In this new ecosystem, digital platforms took on some of the functions of publishers in deciding which content reached consumers. If CompuServe could claim neutrality in its court case, since it didn’t influence what Rumorville published, Facebook has much less of a semblance of neutrality. The algorithm’s curatorial actions are akin to a newspaper choosing which stories to put on the front page. Section 230 has served as a shield, distancing social networks from what individual users post on their platforms. That ranges from #MeToo investigations to racist comments or threats of violence. The law has become the target of growing skepticism, and lawsuits are attempting to hold social networks responsible for what they distribute.

2015年11月，巴黎发生了一系列恐怖袭击，ISIS后来声称对此负责。130人遇难，其中包括一名23岁的美国学生诺希米·冈萨雷斯。冈萨雷斯的家人认为谷歌、推特和脸书（尤其是谷歌旗下的YouTube）应对她的死负责，因为这些公司通过算法推荐向用户推广了ISIS相关内容，从而在煽动巴黎袭击者激进化方面发挥了作用。本质上，他们必须在一定程度上对其平台上的内容以及内容推送给观众的方式负责。该诉讼最初被驳回，理由是谷歌并未通过广告收入为ISIS提供实质性支持，并且对ISIS相关内容的对待与其他内容并无区别，保持了其作为分销商的理论上的中立性。

In November 2015, a series of terrorist attacks, later claimed by ISIS, struck Paris. One hundred thirty people were killed, including a twenty-three-year-old American student named Nohemi Gonzalez. Gonzalez’s family held Google, Twitter, and Facebook responsible for her death—particularly Google’s YouTube—because the companies had promoted ISIS-related content to their users via algorithmic recommendations, and thus played a role in radicalizing the Paris attackers. In essence, they had to be responsible, to a degree, both for what was on their platforms and how it was pushed to viewers. The lawsuit was initially dismissed, on the grounds that Google was not materially supporting ISIS with its ad revenue and did not treat ISIS-related content any differently than other content, maintaining its theoretical neutrality as a distributor.

但在 2022 年 10 月，最高法院决定将此案与另一起针对数字平台的诉讼一并受理，这或许将在我们这个算法无处不在的时代开创一个重要的先例。冈萨雷斯家族在该案的上诉书中辩称：“第 230 条是否适用于这些算法生成的推荐具有巨大的实际意义。交互式计算机服务不断以各种形式向几乎每个使用社交媒体的美国成年人和儿童提供此类推荐。”Facebook 的举报人弗朗西斯·豪根 (Frances Haugen) 也反对该法律，她泄露的文件证明 Facebook 意识到其算法推荐可能造成的损害。豪根在 2021 年告诉参议院小组：“如果我们修改第 230 条，让 Facebook 为其故意排名决策的后果负责，我认为他们会取消基于参与度的排名。”

But in October 2022, the Supreme Court decided to take up the case alongside another lawsuit against digital platforms, perhaps identifying an important precedent to set in our era of omnipresent algorithms. “Whether Section 230 applies to these algorithm-generated recommendations is of enormous practical importance. Interactive computer services constantly direct such recommendations, in one form or another, at virtually every adult and child in the United States who uses social media,” the Gonzalez family argued in the case’s appeal. The Facebook whistleblower Frances Haugen, who leaked documents proving that the company was aware of the damage its algorithmic recommendations could cause, has also argued against the law. “If we reformed Section 230 to make Facebook responsible for the consequences of their intentional ranking decisions, I think they would get rid of engagement-based ranking,” Haugen told a Senate panel in 2021.

算法推送有助于自动传播虚假信息，并可能加速意识形态极端化，向用户推送单一类别中越来越极端的内容。第230条的问题在于，归根结底，奇怪的是，该法律规定目前无人对算法推荐的影响负责。科技公司本身不承担责任，而是系统本身。仅受内部监管，用户只能自生自灭，缺乏基本的内容审核。如果算法推送信息对我们造成不公，或助长了滥用或剥削的网络环境，作为用户和公民，我们几乎无力回天。我们唯一的选择之一就是转向其他平台，但这也已经受到垄断问题的限制。我们与算法推送信息的关系就像一个陷阱：我们既无法影响它们，也无法逃脱它们。

Algorithmic feeds help automatically distribute misinformation and can speed ideological radicalization, feeding users ever more extreme content in a single category. The problem with Section 230 is that ultimately, and strangely, the law makes it so that no one is currently responsible for the effects of algorithmic recommendations. The tech companies themselves aren’t liable, the systems are only regulated internally, and users are left to fend for themselves, short of basic content moderation. If algorithmic feeds mistreat us or contribute to an abusive or exploitative online environment, we, as users and citizens, have little recourse. One of our only possibilities is switching to another platform, and yet that, too, has already been limited by the problem of monopolization. Our relationship to algorithmic feeds feels like a trap: we can neither influence them nor escape them.

废除第230条并非万能药。从某种程度上来说，它保护了用户的网络言论自由，并允许我们所知的数字平台继续存在，而不会因为每一条推文中的侮辱或指控而被起诉。但也有一些修改该条款的提案。2021年10月，在豪根（Haugen）举报Facebook事件后，众议院议员提出了一项名为《反恶意算法正义法案》（JAMA）的法案。该法案将“在网络平台明知故犯地或鲁莽地使用算法或其他技术推荐对身体或严重情感伤害造成实质性影响的内容时，解除第230条的责任保护”。它取消了第230条在任何“基于特定个人信息的个性化推荐”情况下的保护。该法案不涵盖来自用户自身搜索的结果（例如谷歌查询），也不涵盖支撑社交网络的服务（例如网络托管和数据存储），这些服务更加中立，更符合第230条的初衷。

Repealing Section 230 would not be a panacea. In some ways, it protects users’ free speech online and allows digital platforms as we know them to exist without being sued into oblivion for every stray tweeted insult or accusation. But there are proposals to modify it. In October 2021, following Haugen’s Facebook whistleblowing, lawmakers in the House of Representatives introduced legislation called the Justice Against Malicious Algorithms Act, or JAMA. The bill would “lift the Section 230 liability shield when an online platform knowingly or recklessly uses an algorithm or other technology to recommend content that materially contributes to physical or severe emotional injury.” It removes Section 230’s protection in any instance of “personalized recommendations based on information specific to an individual.” The bill does not cover results that come from a user’s own search, like a Google query, or services like web hosting and data storage that undergird social networks, which are more neutral and fit better under Section 230’s original purpose.

2023年2月，最高法院审理了冈萨雷斯诉谷歌案和推特诉塔姆纳案，其中另一起与ISIS有关的恐怖袭击受害者家属根据《反恐法》起诉推特、谷歌和脸书，指控这些公司托管并向用户推荐ISIS内容，从而为该恐怖组织提供有意义的支持。

In February of 2023, the Supreme Court held hearings for the cases of Gonzalez v. Google and Twitter v. Taamneh, in which the family of a victim of another ISIS-linked terrorist attack sued Twitter, Google, and Facebook under the Anti-Terrorism Act, alleging that the companies had hosted and recommended ISIS content to users, thus providing meaningful support to the terrorist group.

早些时候，大法官埃琳娜·卡根（Elena Kagan）几乎开玩笑地指出，法院对数字平台相关问题的无知。“这九位可不是互联网领域最伟大的专家，”卡根说道，引得法庭哄堂大笑。但她也反思了算法信息流的主导影响力：“每当有人在互联网上浏览任何东西时，都离不开算法，”她说。大法官们探讨了算法信息流的用途和能力。建议，并争论算法是否可以被视为“中立”（我认为它们不能），但普遍的不理解显而易见。原告律师埃里克·施纳珀（Eric Sc​​hnapper）在描述YouTube视频缩略图等互联网主流技术时磕磕绊绊。2023年5月，最高法院裁定科技公司不承担责任，并再次维持了对第230条的最强硬解释。

Early on, Justice Elena Kagan all but joked about the court’s ignorance of the issues around digital platforms. “These are not the nine greatest experts on the Internet,” Kagan said, to a wave of laughter in the courtroom. But she also reflected on the dominant influence of algorithmic feeds: “Every time anybody looks at anything on the Internet, there is an algorithm involved,” she said. The justices probed the uses and capabilities of algorithmic recommendations and debated if algorithms can be considered “neutral” (I would argue they cannot), but the general incomprehension was palpable. The plaintiff’s lawyer, Eric Schnapper, stumbled painfully while describing Internet mainstays like YouTube video thumbnails. In May 2023, the Supreme Court ruled that the tech companies were not liable, and upheld the strongest interpretation of Section 230 once again.

根据大幅修改后的《第230条》，我们的信息流将大不相同。社交网络将被迫对其网站上每一条接受算法推广的内容负责。或许他们会调整策略，将大多数内容排除在推荐范围之外，这意味着用户必须刻意关注或搜索特定主题。那些仍然接受算法推广的内容必须经过审查，以确保其完全无关紧要——比如可爱的宠物滑稽动作片段和令人愉悦的新闻报道。选择将取决于集体认知，即什么是可接受的或中性的新闻。（这类内容在某种程度上仍会通过单一过滤器进行过滤。）例如，TikTok 广泛的小众推荐将不复存在。可以想象，TikTok 的信息流完全由适合平淡无奇的电视节目《美国最搞笑家庭录像》的片段组成。它仍然会很有趣，但几乎不会让人上瘾或被操纵。内容审核必须更加严格。

Our feeds under a dramatically reformed Section 230 would look very different. Social networks would be forced to take responsibility for each piece of content on their sites that receives algorithmic promotion. Perhaps they would adapt by putting most content outside the reach of recommendations, meaning that users would have to intentionally follow or search for a given subject. The content that still receives algorithmic promotion would have to be vetted to ensure that it is wholly anodyne—clips of cute pet antics and feel-good news stories. The selection would be subject to a collective perception of what constituted acceptable, or neutral, news. (Such content would still be filtered, in a way, by having a singular filter imposed on it.) TikTok’s wide range of niche recommendations, for example, would be impossible. One could imagine a TikTok feed made up of only clips that would fit in the anodyne TV show America’s Funniest Home Videos. It would still be amusing, but hardly addictive or manipulative. Content moderation would have to be much stricter.

如果能限制有害内容在网络上的曝光，改变这种平衡，强调线性、可选择性内容而非自动推荐，或许是件好事。但这将使互联网更加“干净”，速度也必然会更慢。问题在于，哪些内容应该能够在 Filterworld 上快速流畅地传播，哪些内容应该被放慢速度甚至完全停止。

Changing the balance to emphasize linear, opt-in content over automated recommendations might be a good thing, if it can limit the possible exposure of harmful material online. But this would be a much more sanitized, and by necessity slower, Internet. The problem comes down to determining which kinds of content should be able to travel so quickly and frictionlessly across Filterworld, and which should be slowed down or stopped entirely.

如果说2000年代见证了主流互联网的兴起，2010年代见证了大型数字平台的崛起和主导地位，那么下一个十年似乎将再次拥抱去中心化。“自主权”或许是未来的关键词：个人用户能够自主决定自己发布和浏览内容的方式。我希望互联网能更像Geocities，能够展现个性。以及无处不在的定制化，但多媒体创新也让2020年代的互联网如此引人注目。它将是一个更加混乱、更有趣的地方——更像一个游乐场或沙箱，而不是像互联网现在那样，变成办公室里的格子间。

If the 2000s saw the emergence of the mainstream Internet, and the 2010s saw the rise and domination of massive digital platforms, then the next decade seems likely to embrace decentralization once more. Agency might be the watchword: the ability of an individual user to dictate how they publish and view content. I have hope for an Internet that’s more like Geocities, with expressions of individuality and customization everywhere, but with the multimedia innovations that have made the 2020s’ Internet so compelling. It would be a messier, more fun place—more like a playground or a sandbox than the cubicle office floor that the Internet has come to resemble.

像 Mastodon 这样的开源软件，允许用户创建并托管类似 Twitter 的社交网络，为未来的发展提供了一些线索。但 Mastodon 也展现了这种不同基础设施的一些弊端。自托管平台的受众规模较小，互动也更加困难。你可能找不到任何你想要的内容。既没有病毒式传播的可能性，也没有病毒式传播的威胁。但为了打造更可持续的整体数字文化，我们或许必须做出这些权衡。

Open-source software like Mastodon, which allows its users to create and host their own Twitter-like social networks, provides one hint at what might be to come. But Mastodon also demonstrates some of the disadvantages of such different infrastructure. Audiences are smaller on self-hosted platforms, and interactions are more difficult. You might not be able to find any kind of content that you want. There isn’t the possibility, or threat, of viral fame. But those are the trade-offs that we may have to make for a more sustainable overall digital culture.

减缓放大

SLOWING DOWN AMPLIFICATION

正如英国自杀少女莫莉·拉塞尔在面对铺天盖地的抑郁内容时所经历的那样，推荐对负面内容的加速作用丝毫不亚于正面内容。虽然系统可能对所有内容一视同仁，但推广的后果却并非如此。在新冠疫情期间，网络上错误信息的传播引发了伊维菌素（一种最常用于马匹的药物）的病毒式热潮。服用该药物的患者受到的伤害远大于得到的好处，有时甚至因为服用了这种药物而住院。伊维菌素的故事之所以传播广泛，是因为它们吸引了人们的关注，部分原因是特朗普及其政府的言论将这一问题政治化了。事实证明，这些错误信息针对算法进行了优化，创造了一种即时互动，从而促使更多推广。

As Molly Russell, the British teenager who died by suicide, experienced with the avalanche of depression content, recommendations accelerate negative material as much as positive material. While the system may treat all content the same, the consequences of promotion are not. The spread of misinformation online during the COVID-19 pandemic drove a viral craze for Ivermectin, a drug that is most often used for horses. The patients who took it were hurt much more than helped, sometimes even hospitalized because they ingested the medicine. The Ivermectin stories spread far because they attracted engagement, in part as a politicized issue fueled by the rhetoric coming from Trump and his administration. The misinformation turned out to be optimized for the algorithmic equation, creating the kind of instant interaction that prompts even more promotion.

处理这个问题的常用策略是审核，或者说出于安全考虑的审查：有问题的内容应该被彻底屏蔽，从一开始就不允许出现在用户的动态中。社交网络会通过机器学习分析（例如自动审查特定关键词）和人工审核员（手动决定哪些内容允许通过）进行过滤。Facebook 将大部分人工审核工作外包给了埃森哲是一家名为埃森哲的公司，在全球雇佣了数千名内容审核员，包括葡萄牙和马来西亚等国。这些劳动者每天都要面对镜头下的死亡、虐待录像和儿童色情内容。他们为了我们其他人维护内容的纯净，让自己暴露在精神伤害之下，就像在加纳和其他地方翻找倾倒的国际电子垃圾的拾荒者接触有毒化学物质一样。这些有毒物质不会因为算法的介入而神奇地消失。人类的劳动再次被掩盖。

The usual strategy to deal with this problem is moderation, or censorship for safety’s sake: the problematic content is meant to be blocked out entirely, not allowed into users’ feeds in the first place. Social networks filter both by means of machine-learning analysis—automatically censoring particular keywords, for example—as well as human moderators, who manually decide which content to let through. Facebook outsources much of its human moderation to a company called Accenture, which employs thousands of moderators around the world, including in countries like Portugal and Malaysia. These laborers face daily exposure to deaths on camera, recorded abuse, and child pornography. They keep the feeds clean for the rest of us, exposing themselves to psychic harm the way trash pickers digging through international electronic waste dumped in Ghana and elsewhere are exposed to poisonous chemicals. The toxic material doesn’t just magically vanish because of the mediation of the algorithm. Once again, the human labor is obscured.

然而，有些内容总是会被漏掉，一旦它出现在信息流中，就无法阻止它被推荐。对抗算法越权的一种方法是规范推荐内容的类型，正如第 230 条的一些改革所建议的那样。在英国，一项名为《网络安全法案》的法律正在审议中，旨在防止像 Russell 这样的案件发生。该法案的议会声明规定：“所有可能被儿童访问的服务都有义务保护他们免受有害内容的侵害。”允许此类有害内容的平台还必须公开其在审核和推荐方面的处理方式。这可能包括删除或“降低”有害内容的优先级——即减少推荐频率或根本不推荐。

Some content always falls through the cracks, however, and once it’s in the feed, nothing prevents it from being recommended. One way to combat algorithmic overreach is to regulate which kinds of content recommendations affect, as some reforms to Section 230 suggest. In the United Kingdom, a law called the Online Safety Bill is under consideration, designed to prevent cases like Russell’s. “All services likely to be accessed by children will have duties to protect them from harmful content,” the bill’s Parliamentary statement dictates. Platforms that allow such harmful content will also have to be transparent about how it is treated in terms of moderation and recommendation. This could include removing or “deprioritizing” the harmful content—that is, recommending it less often or not at all.

社交网络必须处理好一系列棘手的优先事项。我们用它们来直接沟通，作为消息服务或关注好友动态的方式。但它们也扮演着广播的角色，触达数百万用户。无论内容是私密的还是公开的，都受制于相同的规则和力量。康奈尔大学技术学者、现任微软首席研究员的塔尔顿·吉莱斯皮解释了这种差距的含义。

Social networks must navigate a difficult set of priorities. We use them to communicate directly with each other, as messaging services or ways to follow updates from our friends. But they also play a broadcast role, reaching audiences of millions of users. Both of those kinds of content are subject to the same rules and forces, whether they are meant to be private or public. Tarleton Gillespie, a technology scholar at Cornell University who is now a principal researcher for Microsoft, explained what that gap means.

“平台为我们提供了某种旨在进行亲密、连贯对话的内容。然后，他们以一种统计上广泛而系统的方式对其进行审核，”吉莱斯皮告诉我。几乎没有空间去考虑特定内容的具体性质，因为它缺乏语境，在整个信息流中被原子化了。这意味着我们也对自己发布的内容如何在网上传播形成了预期——也就是说，我们通常期望它传播得尽可能远。“人们已经形成了一种观念，不仅仅是他们能够“我们只是说说而已，但对于它最终会走向何方，我们也有一定的预期，”吉莱斯皮告诉我。换句话说，我们几乎把算法晋升视为一种权利。

“Platforms have offered us up something that is meant for intimate, connected speech. Then they moderate it in a statistically broad, systematic way,” Gillespie told me. There’s little room for considering the specific nature of a given piece of content, since it lacks context and becomes atomized in the overall feed. That means we have also developed expectations about how the content we post travels online—namely, we most often expect it to travel as far as possible. “People have developed a notion not just that they get to speak, but there’s some expectation of where it’s supposed to go,” Gillespie told me. In other words, we have come to expect algorithmic promotion almost as a right.

但或许这种期望需要改变，我们需要学会在一个没有那么多自动化加速的数字世界中生存。触达广泛的陌生人受众并非权利，而是一项特权，并非每个用户或帖子都应享有。“放大效应”一词描述了算法推荐在内容传播方面所扮演的角色：它就像一个扩音器，将普通的言语转化为呐喊。放大效应是 Filterworld 问题的核心；一种格式总是被放大。通过监管来控制放大效应，可以为生态系统创造更多平衡。

But perhaps this expectation needs to change, and we need to learn to exist in a digital world without so much automated acceleration. Reaching wide audiences of strangers isn’t a right; it’s a privilege that doesn’t need to be possible for every individual user or post. The word amplification describes algorithmic recommendations’ role in spreading content more widely than it would otherwise travel: like a megaphone, it turns regular speech into a shout. Amplification is at the core of Filterworld’s problems; one format gets amplified above all. Using regulation to target amplification could create more balance in the ecosystem.

达芙妮·凯勒是斯坦福大学网络政策中心平台监管项目主任。2021年，凯勒通过哥伦比亚大学发表了一篇题为《放大效应及其不满》的文章。她写道：“放大效应既有好处也有坏处。” 监管的诀窍在于“[利用]好处”——比如帮助用户发现新的声音或兴趣——同时“减少随之而来的危害”，比如加速问题内容的传播。“在美国，大多数关于算法监管的讨论都与内容有关，”凯勒告诉我。但“对于哪些内容应该呈现给人们，并没有一个基准。”我们可以抱怨在Spotify上听到太多相同的音乐，或者在Facebook上看到太多家人的政治观点，但没有一种完美的内容组合、主题或语气选择能够适合所有人。

Daphne Keller is the director of the Program on Platform Regulation at Stanford’s Cyber Policy Center. In 2021, Keller published an essay through Columbia University titled “Amplification and Its Discontents.” “Amplification features can do both harm and good,” she wrote. The trick in regulation is to “[harness] the benefits”—like helping users discover new voices or interests—while “reducing the attendant harms,” like the acceleration of problematic content. “Most discussions in the U.S. of regulating algorithms are about content,” Keller told me. But “there’s no such thing as a baseline of what content should be surfaced to people.” We can complain about hearing too much of the same music on Spotify or seeing too many of our family members’ political opinions on Facebook, but there’s no one perfect mixture of content, a selection of topics or tones, that fits for everyone.

像 Facebook 这样的算法信息流的构成，可以比作健康饮食的食物金字塔。美国农业部建议，脂肪、油和甜食的摄入量应尽量少，水果和蔬菜的摄入量则要大得多。监管信息流的放大效应，可以迫使信息流按特定比例包含不同类型的内容，将少量淫秽内容与大量信息丰富、政治中立或具有地域特色的内容混合在一起。这样的规则将限制我们目前在信息流中体验到的无限扩展。

The makeup of an algorithmic feed like Facebook’s could be compared to the food pyramid for a healthy diet. The USDA recommends a minimum of fats, oils, and sweets, and a much larger number of fruits and vegetables. Regulating amplification could force feeds to incorporate specific proportions of different kinds of content, mingling a little bit of salacious material with a higher volume of informative, politically neutral, or geographically local material. Such rules would limit the infinite expanse that we now experience in our feeds.

针对儿童电视已经存在类似的法律。《儿童电视法案》于1990年由美国国会通过，并于1997年得到进一步严格。该法案规定，所有主要广播电视台每周必须播放至少三个小时的儿童节目。除了“教育性和信息性”之外，这些电视节目还必须满足其他要求，例如限制广告，不得显示商业网站地址。根据媒体教育中心1997年的一本小册子，该法案的压力促使了《比克曼的世界》和《比尔·奈，科学小子》等节目的制作。

A similar kind of law already exists for children’s television. The Children’s Television Act was passed by Congress in 1990 and made stricter in 1997. It mandated that all major broadcast television stations had to feature at least three hours per week of material directed at kids. Aside from being “educational and informational,” the television programs must meet other requirements, like limiting advertising and not displaying commercial website addresses. Pressure from the law prompted the production of shows like Beakman’s World and Bill Nye, The Science Guy, according to a 1997 pamphlet from the Center for Media Education.

但随着电视从少数几个基础频道分裂成更大的有线电视套餐，再到2010年代不受相同监管约束的流媒体服务，这项法律的适用难度也随之加大。什么样的节目才算真正具有教育意义，也有些模糊；一些频道甚至开始利用面向青少年的真人秀节目来满足他们的教育需求。无论节目本身是否具有教育意义，它仍然可能成为鼓励消费主义的商业手段：《巴尼与朋友》（Barney & Friends）虽然多年来一直由公共电视台制作，但却售出了数百万个品牌玩具。

But the law became more difficult to apply as television splintered away from just a handful of basic channels into larger cable packages and then streaming services in the 2010s, which weren’t subject to the same regulations. What kinds of shows qualify as properly educational is also somewhat ambiguous; some channels even began to use reality television shows marketed toward teenagers to fulfill their educational requirement. No matter if a show was informative, it could still become a commercial enterprise that encouraged consumerism: Barney & Friends, though it was produced for many years by public television, sold millions of branded toys.

《儿童电视法》虽然存在缺陷，但已经持续了数十年，并被证明具有一定的约束力。2007年，美国电视网Univision因违反该法而面临美国联邦通信委员会（FCC）2400万美元的罚款。Univision曾试图将其电视肥皂剧定义为符合教育意义的内容。美国政府不同意，这些节目被证明不符合教育意义，Univision最终以Planeta U的名义重新推出了儿童节目板块。该法案开创了先例，将向观众播放的内容类型进行有意义的影响。如果我们接受这样一个事实：如今的社交网络与电视台等媒体企业一样做出编辑决策，那么它们是否应该被迫时不时地播放某些类型的内容，优先播放那些可能对我们有益——或者最终对社会有益——的信息？

Flawed as it is, the Children’s Television Act has persisted for decades and has proven to have some teeth. In 2007, the American channel Univision faced a $24 million fine from the FCC for breaking the law. It had attempted to define its telenovela soap operas as proper educational content. The US government disagreed, the shows did not prove qualified, and Univision eventually relaunched its children’s programming block under the title of Planeta U. The law creates a precedent for meaningfully influencing what kinds of content are broadcast to viewers. If we accept the fact that today’s social networks make the same kind of editorial decisions as a media business like a TV channel, should they then be forced to feature certain kinds of content every so often, prioritizing information that might be good for us—or ultimately good for society?

算法推荐会考虑到我们潜意识里倾向于的内容：任何我们长时间悬停在鼠标上方的内容，更不用说点击了。当我们在滑动跳过TikTok视频之前犹豫不决时，这些数据就会被考虑进去。就像垃圾食品或成瘾物质一样，人类可能只需要帮助他们选择消费的内容组合。正如凯勒告诉我的那样：“我们与猴子大脑产生共鸣，就像我们在杂货店收银台排队买糖果一样。” 算法推送不仅在个人层面，而且在社交网络所有用户的群体层面加速了这些最糟糕的冲动。挑逗性的内容——可能是暴力、挑衅或误导性的内容——可能比那些更无聊但更有价值的内容更容易被发现。

Algorithmic recommendations take into account what we subconsciously gravitate toward: any piece of content that we mouse over for an extended period of time, let alone click. When we hesitate before swiping to skip a TikTok video, that data is considered. Like junk food or addictive substances, humans might simply need help with choosing the mixture of content they consume. As Keller told me: “We click with our monkey brains, the same ones that cause us to buy a candy bar in the checkout line at the grocery store.” Algorithmic feeds accelerate these worst impulses, not just on an individual level but an aggregate one, across all the users of a social network. Titillating material—content that might be violent, provocative, or misleading—can be easier to discover than material that is more boring but also more valuable.

“这不是构建我们信息饮食的好方法。它会导致社会问题，”凯勒继续说道。她将可能的答案描述为“强迫平台在用户要求的全糖果饮食中加入一些素食”。本质上，我们必须接受算法推荐无法精准地提供我们想要的内容或我们最有可能参与的内容——就像我们接受人类新闻编辑最了解向受众传播什么才是最重要的一样。（我们并不指望《纽约时报》的头版会迎合我们的个人喜好。）信息流中的这些“素食”可以是来自经过审查的出版物列表的故事，也可以是更像公益广告的帖子——这些话题是我们共同认定值得提出的，尽管这种全社会共识在如今似乎比以往任何时候都更加稀缺。或许关于好莱坞电影的更新足够中立，或者国家新闻摘要中穿插着令人愉悦的本地新闻。当然，这种混合正是传统媒体所提供的，它与我们习惯从社交网络获得的内容相反。

“That’s not a great way to structure our information diet. It contributes to societal problems,” Keller continued. She described the possible answer as “forcing platforms to put some veggies in with the all-candy diet that users are asking for.” In essence, we would have to be okay with algorithmic recommendations not giving us precisely what we want or what we would be most likely to engage with—just like we accept that human news editors know best what’s most important to broadcast to their audiences. (We don’t expect the front page of The New York Times to conform to our personal preferences.) Such “veggies” in the feed could take the form of stories from a list of vetted publications or posts that are more like public service announcements—topics that we have collectively decided are worth bringing up, though such society-wide agreement seems scarcer these days than ever. Perhaps updates about Hollywood films are neutral enough, or digests of national news interspersed with feel-good local stories. Of course, that kind of mixture is exactly what traditional media used to provide, and it’s the opposite of what we are accustomed to getting from social networks.

社交媒体的兴起为文化和娱乐创造了一系列新的动态。用户在特定时刻拥有了更多消费选择，创作者只需将内容上传到互联网，就能更轻松地触达受众。我们不必只观看制作人选择在有线电视上播放的内容。我们开始期待个性化，无论是由我们自己的行为还是算法驱动。但这种看似更民主、等级更低的动态也让我们感到，旧的法律法规不再适用，正是因为我们可以决定何时观看或收听某些内容，以及何时做出选择。另一个来源。我们可能拥有更多的独立性，但最终作为消费者，我们得到的保护却更少。

The rise of social media has created a new set of dynamics for culture and entertainment. Users have much more choice of what to consume at a given moment, and creators have a much easier time reaching audiences by simply uploading their content to the Internet. We don’t have to just watch what a producer elects to put on cable television. We have come to expect individualization, whether driven by our own actions or an algorithm. But that seemingly more democratic and low-hierarchy dynamic has also given us a sense that the old laws and regulations don’t apply, precisely because we can decide when to watch or listen to something and when to choose another source. We might have more independence, but we ultimately have less protection as consumers.

监管算法推荐很快就会演变成言论自由问题。正如圣克拉拉大学法学教授、高科技法律研究所联合主任埃里克·戈德曼告诉我的那样：“算法只是一种规范服务言论选择的方式。” 凯勒区分了问题内容（以暴力威胁或仇恨言论为特征）和“合法但具有挑衅性的言论”。如果用户不断点击这类内容，表达对它的渴望，那么法律应该介入吗？在莫莉·拉塞尔的案例中，她接触了关于抑郁症的内容，并获得了更多此类内容的推荐。这本可以是一次积极的体验，帮助她在网上找到社群和同情。但它也变成了负面体验，因为这类内容被推荐得太频繁了，这是算法同质化恶性循环的另一个后果。

Regulating algorithmic recommendations quickly becomes a free speech issue. As Eric Goldman, a professor of law at Santa Clara University who codirects the High Tech Law Institute, told me, “An algorithm is just a way of codifying the service’s speech choices.” Keller made a distinction between problematic content—marked by threats of violence or hate speech—and “speech that’s legal but provocative.” If a user constantly clicks on that kind of content, expressing a desire for it, then should the law stand in the way? In Molly Russell’s case, she engaged with content about depression and was recommended more of it. That could have been a positive experience, helping her to find community and sympathy online. But it also became negative, since such content was recommended too often, another consequence of the vicious cycle of algorithmic homogeneity.

如果要规范推荐，就必须根据内容做出某些决定。“哪些偏好是不可接受的，必须由法律推翻？”凯勒问道。“假设平台不再提供人们想要的东西；你必须给他们其他东西。哪些是蔬菜？我们强行塞进饮食中的哪些信息？我们如何决定这些信息是什么？我不知道答案是什么，”凯勒说。她建议建立一个“内容中立”的“熔断机制”。与其决定哪些内容是正面的或负面的，不如限制任何通过推荐快速传播的内容，而不是进一步推广，从而减缓其传播速度，让审核人员有更多时间判断其是否合适，以免其接触到大量受众。熔断机制还可以让我们回到一个不那么全球化的媒体生态系统，那时内容会更牢固地停留在其原始语境中。虽然这可能会限制像搞笑视频这样无害的内容，但这也意味着拉塞尔可能不会接触到许多伤害她的模因。她就不会收到 Pinterest 上那封催促她进一步陷入困境的“抑郁症 Pins”电子邮件。

If recommendations are to be regulated, certain decisions have to be made based on content. “Which preferences are unacceptable and must be overridden by law?” Keller asked. “Say platforms stop giving people what they want; you have to give them something else instead. What are the veggies, what information are we forcing into the diet, and how do we decide what that is? I don’t know what the answer is,” Keller said. She has suggested a system of “circuit breakers” that are “content neutral.” Rather than deciding which content is positive or negative, any kind of content that goes viral—accelerating quickly via recommendations—would be limited instead of promoted more, slowing down its spread and giving moderators more time to judge if it’s appropriate before it reaches massive audiences. Circuit breakers could also bring us back to a less globalized media ecosystem, when pieces of content stayed more firmly within their original contexts. While this might limit innocuous content like funny videos, it also means that Russell might not have been exposed to many of the memes that harmed her. She wouldn’t have gotten that email of Pinterest “Depression Pins” urging her deeper down the rabbit hole.

关于内容的决定最终必须由人类来做，这既是因为机器学习系统尚无法做出如此细微的区分，也因为其后果会影响人类生活。人类的能力不像软件那样可扩展——一个人无法保护数百万用户，也无法审查每一条通过算法推广的帖子。而危险内容的界定，例如禁用词列表，需要随着文化的变化而不断更新，因为文化从来都不是一成不变的。但还有其他方法可以解构“过滤世界”（Filterworld），这些方法较少依赖于评估内容本身，而更多地关注2010年代互联网的基本结构。如果我们改变平台的运作方式，那么我们最终也能改变体验平台的方式。

Decisions about content ultimately have to be made by humans, both because machine-learning systems are not yet capable of making such subtle distinctions and because their consequences impact human lives. That human capacity is not scalable in the way that software is—one person can’t safeguard millions of users or vet every post that gets algorithmic promotion. And the scope of what qualifies as dangerous, like a list of banned words, requires constant updating as it changes, since culture is never static. But there are other ways of deconstructing Filterworld that rely less on evaluating the content itself and more on targeting the basic structure of the 2010s’ Internet. If we change how the platforms are able to work, then we can also change the end result of how we experience them.

欧盟战略

THE EUROPEAN UNION STRATEGY

在2010年代，Facebook就像一个小侦探，追踪你在互联网上的足迹，记录你浏览和点击的所有内容、搜索的每一个词条，以及与你联系的每一个人。它不仅在Facebook上追踪你，还会在未经你明确许可的情况下，借助追踪cookie和“点赞”按钮，在其他网站上追踪你。它的目标是收集足够的数据，以便在其平台上为你提供个性化推荐和精准投放的广告，将你的注意力出售给那些正在寻找目标人群的广告商，比如那些居住在明尼苏达州、年龄在40到50岁之间、有意购买园艺设备的人。大多数应用程序都以同样的方式运作，构建了一个无所不包的数字监控网络。无论你走到哪里，你的踪迹都会被某个实体获取，慢慢积累成详细的个人资料，供企业利用。逃避追踪的唯一方法是躲避追踪，使用隐身浏览器或通过虚拟专用网络（VPN）伪造身份。即便如此，如果您没有登录帐户，数字平台通常提供的功能较少——当然，这会受到持续跟踪。

Facebook, over the 2010s, was like a tiny detective tracing your steps around the Internet, noting down everything you looked at and clicked on, every term you searched, and every other person to whom you were connected. It didn’t just track you on Facebook itself; without asking for particularly explicit permission, it followed you onto other websites too, with the help of tracking cookies and Like buttons. Its goal was to collect enough data so that it could serve you personalized recommendations and precisely targeted ads on its platform, selling your attention to advertisers who were seeking, say, anyone between the ages of forty and fifty who lived in Minnesota and had shown an interest in buying gardening equipment. Most apps functioned in the same way, creating an all-encompassing net of digital surveillance. Everywhere you went, your trail was picked up by some entity or another, slowly accreting into detailed profiles that the businesses could use to their benefit. The only way to escape the tracking was to hide from it, using an incognito web browser or faking your Identity with a virtual private network. Even then, digital platforms often offered fewer features if you weren’t logged in to an account—which, of course, was continuously tracked.

2016年4月，欧盟通过了一项名为《通用数据保护条例》的法律。该条例自2012年以来一直在制定中，旨在赋予互联网用户对所有在线个人数据的更大权利，并建立统一的监管结构。在欧盟各成员国范围内。GDPR 通常缩写为 GDPR，它使用“数据主体”一词来描述我们所有用户。它指的是任何可通过在线数据识别的人，无论是姓名、位置，还是“特定于该自然人身体、生理、遗传、心理、经济、文化或社会身份的一个或多个因素”。数据主体可能是法律术语，但它是一个令人回味的短语。在哲学中，“主体”是任何具有能动性和独特个人经历的实体。数据似乎与此相反；它是记录经历的非物质、无生命的东西——对已经发生的事情的记录。

In April 2016, the European Union adopted a law called the General Data Protection Regulation. It had been in development since 2012, with the aim of giving Internet users stronger rights to all of that personal data online and creating a unified regulatory structure across the EU’s member states. GDPR, as it’s usually abbreviated, uses the term “data subject” to describe all of us users. It means anyone who is identifiable from their data online, whether their name, location, or “one or more factors specific to the physical, physiological, genetic, mental, economic, cultural or social identity of that natural person.” Data subject might be legalese, but it’s an evocative phrase. In philosophy, a “subject” is any entity with agency and a unique personal experience. Data seems to stand in contrast to that; it’s an immaterial, inanimate thing that documents experience—a record of something that has happened.

GDPR 承认，如今我们就是数据——数据既记录了我们做过的事，也影响着我们能够做的事情，或者未来最有可能做的事情，而这些事情往往是通过算法决策做出的。因此，我们应该对数据拥有一些与我们对自身身体相同的控制权和权利。该法案概述了一系列“数据主体权利”，例如基本人权，数字平台必须遵守这些权利。第一项是透明权，要求公司以“清晰明了的语言”回应用户关于其数据使用方式和原因的信息请求。第二项是透明权，确保用户可以提出此类请求，赋予他们“访问权”，了解数据收集的形式、跟踪发生的时间以及数据存储时长，以及请求获取数据副本的权利。

GDPR recognizes that these days, we are our data—data both documents what we have done and influences what we are able to do, or are most likely to do, in the future, oftentimes through algorithmic decisions. Thus, we should have some of the same kinds of control over it, and rights to it, that we have over our physical bodies. The law outlines a list of “data subject rights,” like fundamental human rights, that digital platforms have to comply with. The first is the right to transparency, forcing companies to respond to users who request information about how and why their data is being used in “clear and plain language.” The second ensures that users can make that request, giving them a “right of access” to information about which forms of data are collected, when tracking happens, and how long data is stored, as well as to request copies of the data itself.

2017年， 《卫报》记者朱迪思·杜波泰尔（Judith Duportail）利用《通用数据保护条例》（GDPR）要求获取Tinder持有的关于她的所有数据——总计八百页，包括她的Facebook点赞记录、平台上每一次匹配和对话的元数据，以及超过一千七百条信息。（她可是个不折不扣的数据主体。）杜波泰尔写道：“我对自己自愿披露的信息量感到惊讶。”但她本不该如此：这些数据正是Tinder产品本身的动力，而自我披露是我们为了自动化效率而做出的牺牲。浪漫的配对——另一种算法推荐——需要深入的了解。

In 2017, Judith Duportail, a journalist working for The Guardian, used GDPR to request all of the data that Tinder had on her—which amounted to eight hundred pages, including her Facebook likes; metadata about every match and conversation she had on the platform; and over seventeen hundred messages. (She was quite a data subject.) Duportail was “amazed by how much information I was voluntarily disclosing,” she wrote. But she shouldn’t have been: such data is the fuel for Tinder’s product itself, and self-disclosure is the trade-off we make for automated efficiency. Romantic matchups—another kind of algorithmic recommendation—require intimate knowledge.

GDPR 的第三组权利涉及修改数据本身。“更正权”是指能够编辑或更正关于您自己的数据，而“删除权”则确保如果个人数据“不再需要”、用户撤回同意或非法收集，则将被删除。这项权利也被更诗意地称为“被遗忘权”，欧盟自2014年以来就已存在这项权利，但美国没有类似的权利。第四项权利是关于选择退出的，赋予用户“反对权”，选择不再被追踪。这尤其适用于广告：“如果数据主体反对为直接营销目的而进行的数据处理，则个人数据不得再为此目的进行处理，”法律规定。如果您不想被追踪用于定向广告，您可以不必这样做——尽管广告当然可以通过其他方式联系到您。它还包括“不受仅基于自动化处理的决策约束的权利”。换句话说，除非用户同意，否则他们无需经历任何算法操作。您可以阻止Facebook的小侦探跟踪您，以弄清楚接下来该提出什么建议。

GDPR’s third group of rights deals with modifying the data itself. The “right to rectification” means being able to edit or correct the data about yourself, while the “right to erasure” ensures that personal data will be deleted if they “are no longer necessary,” if the user withdraws consent, or if they are unlawfully collected. It’s also known more poetically as the “right to be forgotten,” and it has existed in the European Union since 2014, though the United States has no equivalent. The fourth group is about opting out, giving users a “right to object,” to choose to no longer be tracked. This is particularly applicable to advertising: “Where the data subject objects to processing for direct marketing purposes, the personal data shall no longer be processed for such purposes,” the law dictates. If you don’t want to be tracked for the sake of targeted advertising, you don’t have to be—though ads can certainly reach you in other ways. It also includes “the right not to be subject to a decision based solely on automated processing.” In other words, a user need not experience any algorithmic action unless they consent to it. You can prevent Facebook’s tiny detective from following you around to figure out what to suggest next.

GDPR于2018年5月25日生效，影响范围不仅局限于欧洲公司，还包括所有向欧盟公民提供商品或服务的公司，几乎涵盖所有主流数字平台。表面上看，变化似乎不大。网站突然弹出窗口，在您首次访问时请求允许追踪您，并提供机会仔细阅读详尽的服务条款文件。“Cookies”——网站用来追踪您的数据包的术语，其名称源于20世纪90年代计算机时代的幸运饼干——已成为常用语。但它在追究网站对肆意积累个人数据的责任和执行实际后果方面代表着翻天覆地的变化。在缓慢起步之后，欧盟国家一直在积极将该法应用于其管辖范围内的公司，从2020年开始，罚单数量持续增长。截至2023年初，根据GDPR开出的罚款超过1300张，总额超过23亿欧元。

GDPR went into effect on May 25, 2018, impacting not just companies based in Europe but any company that offers goods or services to EU citizens, which includes almost every major digital platform. On the surface, the changes didn’t look like much. Websites suddenly implemented pop-ups that, on your first visit, asked for permission to track you and offered the chance to peruse elaborate terms of service documents. “Cookies”—the term for the packets of data that websites use to track you, which derived their name from fortune cookies in 1990s computing—became common parlance. But it represented a sea change in terms of holding websites accountable for the wanton accumulation of personal data and enforcing actual consequences. After a slow start, EU countries have been actively applying the law to companies based in their jurisdictions, with a consistent growth in citations beginning in 2020. As of early 2023, there were over thirteen hundred fines under GDPR, totaling over 2.3 billion euros.

2022年11月，Facebook因违反GDPR面临2.75亿美元的罚款，此前该公司发生数据泄露事件，导致超过五亿用户的个人信息被泄露，并被传播到一个黑客论坛上。这并非该公司第一次被罚款。去年9月，Instagram因未能保护未成年人数据而再次违规，导致其损失4亿美元。2021 年，WhatsApp 因缺乏明确的隐私政策被罚款 2.25 亿欧元。罚款金额最高的是亚马逊，它因跟踪用户数据且不允许用户选择退出而被罚款 7.46 亿欧元。除了科技巨头，其他公司也违反了法律。2020 年 3 月，荷兰数据保护局对荷兰皇家草地网球协会（荷兰网球的国家管理机构）处以 52.5 万欧元的罚款，原因是“非法出售个人数据”。该协会在未事先征得其同意的情况下，将其 35 万名会员的数据出售给赞助商用于营销目的，违反了 GDPR 的规定，即此类交易必须是选择加入的。

In November 2022, Facebook faced a $275 million fine for violating GDPR, after a data leak exposed the personal information of over five hundred million users, which was disseminated on a hacker forum. It wasn’t the company’s first fine. A September violation cost it $400 million after Instagram failed to protect the data of minors using its service, and in 2021, WhatsApp was fined 225 million euros for lacking clear privacy policies. The largest fine went to Amazon, which was charged 746 million euros for tracking users’ data and not allowing them to opt out. Companies other than tech giants violate the law, too. In March 2020, the Dutch Data Protection Authority imposed a fine of 525,000 euros on the Royal Dutch Lawn Tennis Association, the national governing body for tennis in the Nether–lands, for “an illegal sale of personal data.” The association had sold the data of its 350,000 members to its sponsors for marketing purposes without first asking for their consent—breaking GDPR’s mandate that such a transaction has to be opt-in.

虽然这些罚款与科技公司的年收入相比微不足道，但它们确实表明了GDPR如何能够促使一定程度的合规。虽然数据权利法案听起来像是Filterworld的理想解决方案，但该法律在其他方面却令人失望。只需点击“接受所有Cookie”按钮，用户就会像以前一样被追踪。正如美国倡导组织Accountable Tech的联合创始人妮可·吉尔（Nicole Gill）所观察到的，让被追踪的选择尽可能无缝和被动符合公司的最佳利益。“在线服务会找到一种对用户来说摩擦最小的遵守法律的方式，”吉尔说道。无摩擦始终是Filterworld的理想——一旦你放慢速度，你可能会重新考虑你点击的内容，并泄露你的数据。“摩擦让人们思考自己的行为，”她继续说道，这一点同样适用于Spotify电台或TikTok动态。如果你想得太多，你可能会停下来。

Though these penalties pale in comparison to tech companies’ annual revenue, they do demonstrate how GDPR can provoke a certain amount of compliance. While a bill of data rights sounds like an ideal solution to Filterworld, the law has been disappointing in other ways. All it takes is the click of an “accept all cookies” button for a user to be tracked just as much as they were before. As Nicole Gill, the cofounder of an American advocacy group called Accountable Tech, observed, it’s in the companies’ best interests to make choosing to be tracked as seamless and passive as possible. “Online services will find a way to follow the law that offers the least amount of friction to their users,” Gill said. Frictionlessness is always the Filterworld ideal—as soon as you slow down, you might just reconsider what you’re clicking on and giving your data away. “Friction allows people to think about their actions,” she continued, a point that applies just as well to Spotify radio or the TikTok feed. If you think too much, you might stop.

我自己也对这种被动感到内疚。当类似的GDPR通知开始出现在美国网站上时，我通常只是点击一下，就放弃了我的数据。如果我喜欢阅读某个网站——比如全国性食品刊物Eater——我会欣然接受追踪，因为我觉得（或许是错误的）我可以信任这个网站。当《卫报》（我欣赏它，认为它是一份英国左翼报纸）向我索要数据时，我也答应了。这真的会有什么问题吗？不得不选择退出感觉就像是一种劳动。如果我感到自己是正义的，我会愤怒地点击“不”。我当然不想要这种监控！但如果我感到愤世嫉俗，我会接受，因为也许它会改善我的算法推荐并带来一些引人注目的个性化。

I’m guilty of this passivity myself. When such GDPR notices began popping up on American websites, I most often just clicked to give away my data. If there was a website I liked to read—Eater, for example, the national food publication—I accepted the tracking readily, because I figured, perhaps wrongly, that I could trust the site. When The Guardian, which I appreciated as a left-leaning British newspaper, asked me for my data I said yes, too. What could really go wrong? Having to opt out felt like a form of labor. If I was feeling righteous, I would angrily hit No. Of course I don’t want this surveillance! But if I was feeling cynical, I would just accept, because maybe it would improve my algorithmic recommendations and result in some compelling personalization.

部分原因是出于懒惰和界面设计的技巧。“选择加入”按钮通常比“选择退出”按钮颜色更深、更显眼，所以我花了很长时间才分清哪个是哪个。但我个人的选择似乎并不一定会改变网站的功能，无论是对我还是对其他人而言。“大多数人都会接受，”阿姆斯特丹大学数字平台学者、前斯坦福互联网与社会研究中心研究员帕迪·莱尔森 (Paddy Leerssen) 说道。最终，这项法律可能会将负担更多地放在用户身上，而不是公司身上。“GDPR 创建的这种个人责任机制实际上并不有效，”莱尔森继续说道。欧盟的新法律不是针对数据，而是更具体地针对推荐系统。莱尔森认为，这些是“命令与控制型法规，政府在告诉行业该做什么，而不是交给用户自行选择。”

In part it was out of laziness and the tricks of interface design. The opt-in button is often darker and more prominent than the opt-out, so my brain took a beat too long to understand which was which. But it also felt like my personal choice wouldn’t necessarily change how the website functioned, for me or anyone else. “Most people are just accepting anyway,” said Paddy Leerssen, a scholar of digital platforms at the University of Amsterdam and a former fellow at the Stanford Center for Internet and Society. Ultimately, the law might place the burden more on the user than on the company. “This whole kind of individual responsibility type mechanism that the GDPR creates isn’t really effective,” Leerssen continued. Rather than targeting the data, newer EU laws take action against recommender systems more specifically. According to Leerssen, they are “command and control regulations, where the government is telling the industry what to do, rather than leaving it to a matter of user choice.”

《数字服务法案》于2022年7月获得批准，并将于2024年生效。该法案规定，在推荐方面，与GDPR在数据方面所做的规定类似，也要求透明度和沟通：平台“应以易于理解的方式，清晰地呈现此类推荐系统的主要参数，以确保接收者了解信息的优先级。” 但该法案也规定，算法信息流必须可定制，使用户能够随意更改变量的平衡，或选择完全不使用个人数据的信息流：“这些选项不基于对接收者的分析。”

The Digital Services Act, which was approved in July 2022 and goes into effect in 2024, provides for some of the same kinds of transparency and communication around recommendations that GDPR does for data: Platforms “should clearly present the main parameters for such recommender systems in an easily comprehensible manner to ensure that the recipients understand how information is prioritized for them.” But it also says that algorithmic feeds must be customizable, enabling users to change the balance of variables at will or choose a feed that doesn’t leverage personal data at all: “options that are not based on profiling of the recipient.”

另一项欧盟法律《数字市场法》于2022年9月签署成为法律，旨在解决垄断问题，鼓励竞争。科技巨头被称为“守门人”。该法禁止未经用户同意，合并同一家公司运营的不同服务（例如Meta旗下的Facebook和WhatsApp）的数据。它还禁止“自我偏好”，即谷歌和亚马逊有时会以中立的自动推荐为幌子推广自家产品，例如在搜索结果中——这是强化当前互联网同质化的伎俩之一。DMA罚款最高可达公司年收入的 10%，对于屡次违法者，罚款最高可达 20%，因此合规变得更加重要。

Another EU law, the Digital Markets Act, was signed into law in September 2022 and addresses monopolization, encouraging competition. The tech giants are labeled “gatekeepers.” The law bans combining data from different services operated by one company, such as Meta’s Facebook and WhatsApp, without the user’s consent. It also prohibits “self-preferencing,” the way that Google and Amazon at times promote their own products under the guise of neutral automated recommendations, as in search results—one of the tricks that reinforces the homogeneity of the current Internet. DMA penalties can run up to 10 percent of a company’s annual revenue, and 20 percent for repeat offenders, making compliance even more vital.

这些法律生效后，很可能会彻底改变我们的算法格局，赋予用户在推荐和内容推送配置方面更大的自主权。随着我们开始了解自己的偏好，并根据自己的品味塑造数字生活，这种被动的关系将变得更加主动。算法推送将不再像现在这样单一和难以理解，而更像是一种功能性工具。你的推送没必要和我的一模一样。由此产生的丰富性也可能带来网络文化的更多元性。

As they go into effect, these laws are likely to overhaul our algorithmic landscape, giving users much more agency when it comes to recommendations and the configuration of a content feed. The passive relationship would become a more active one as we begin to figure out our own preferences and shape our digital lives to follow our own tastes. Algorithmic feeds will appear less monolithic and impenetrable, as they do now, and more like the functional tools they are. There’s no reason your feed needs to work exactly like my feed. The resulting profusion could lead to more diversity of culture online, as well.

科技公司正在应对新的法律环境。2023年8月，母公司Meta宣布Facebook和Instagram都将为用户提供完全退出算法推荐的选项，从而消除自动个性化的可能性。但该选项仅适用于欧盟用户——因为美国在通过此类立法方面进展缓慢。看到这些头条新闻时，我感到羡慕不已。突然之间，仿佛只有欧盟居民才能呼吸到无污染的空气。

Tech companies are responding to the new legal landscape. In August 2023, the parent company Meta announced that both Facebook and Instagram would add options for users to opt out of algorithmic recommendations entirely, removing the possibility of automated personalization. But that option would only be available for users in the EU—because the US has been much slower to adopt such legislation. When I saw the headlines, I was jealous. It was suddenly as if only EU residents could breathe pollution-free air.

美国法规

AMERICAN REGULATION

欧盟数字平台监管已预示着美国社交网络将面临重大变革。尽管这些法律不一定适用于美国，但它们已形成罕见的变革压力。2021年4月，苹果宣布对其iPhone操作系统进行一项名为“应用追踪透明度”的更新。所有想要追踪用户数据用于广告目的的应用都必须先获得许可，并在弹出窗口中允许用户选择退出追踪。苹果还新增了一个追踪菜单，用户可以选择关闭所有已下载应用的追踪功能。这项改变可能看起来并不显著，但它的后果却立竿见影：早期统计数据显示，只有16%的用户选择启用追踪功能，但一年后这一比例上升到了25%左右。由于我的手机比网页浏览器更具私密性，所以我通常会选择退出。至少在移动设备上，这项功能彻底摧毁了这种定向广告构成了科技公司的大部分收入。2022年初，Facebook预测该功能将使其损失高达100亿美元，这导致其股价下跌超过26%，损失超过2300亿美元。

EU digital platform regulation has already heralded a major disruption in American social networks. Even though the laws don’t necessarily apply in the United States, they have created a rare pressure to change. In April 2021, Apple announced an update to its iPhone operating system called App Tracking Transparency. Every app that wanted to track its users’ data for advertising purposes would have to ask permission first, in a pop-up that allows users to opt out of being tracked. Apple also added a tracking menu with an option to turn off tracking from every downloaded app. It may not have looked like a significant change, but its consequences were immediate: early statistics showed that only 16 percent of users opted in to tracking, though that rate increased to around 25 percent a year later. Because my phone felt more personal and intimate than a web browser, I usually opted out. At least on mobile, the feature kneecapped the kind of targeted advertising that makes up the majority of tech companies’ revenue. In early 2022, Facebook predicted that it would lose as much as $10 billion from the feature, which caused its stock price to drop more than 26 percent—a loss of more than $230 billion.

苹果的应用追踪透明度计划 (App Tracking Transparency) 为美国用户提供了一个现实世界的实验，看看他们如何应对类似《通用数据保护条例》(GDPR) 提供的强制数据隐私保护。事实证明，除非我们意识到被追踪的好处，否则我们并不特别希望被追踪。（例如，一些电子游戏应用的“选择加入”追踪比例就高得多。）苹果通过提供简单的隐私功能脱颖而出，并在此过程中打击了竞争对手，这表明，即使是最大的公司也并非像表面看起来那么无懈可击。数据访问是他们的致命弱点；切断他们收集数据的能力，就能打破 Filterworld 的控制。

Apple’s App Tracking Transparency presented a real-world experiment in how American users would react to enforced data privacy like the kind provided by GDPR. It turns out that we don’t particularly want to be tracked unless we see the benefits of doing so. (Some video-game apps had much higher rates of opt-in tracking, for example.) Apple differentiated itself by offering simple privacy features and hurt its competitor in the process, demonstrating that perhaps the largest companies aren’t so invulnerable as they seem. Data access is their Achilles’ heel; cutting off their ability to collect it breaks the grip of Filterworld.

斯坦福大学法学教授、斯坦福网络政策中心联席主任纳撒尼尔·佩西利（Nathaniel Persily）等人士表示，欧盟立法的形式与美国立法通常有所不同。他告诉我，包括《通用数据保护条例》（GDPR）和《数据安全法案》（DSA）在内的欧盟法律往往范围更广、更模糊，要求采取的行动可能无法立即实施。这些法律“需要几十年的时间才能完善”。但此类法规将为其他国家提供一个可行的模式和前进的道路。“在这里，欧洲将充当美国摇尾巴的角色，”佩西利说道。《通用数据保护条例》的持续存在也证明了另一个重要论点：加强个人数据保护并不会破坏互联网。我们可以像以前一样消费更多内容，只是安全感和自主性更强了。

EU legislation arrives in a different form than US lawmaking usually does, according to Nathaniel Persily, a professor of law at Stanford University and the codirector of the Stanford Cyber Policy Center, among other groups. He told me that EU laws, including GDPR and DSA, tend to be broader and more ambiguous, demanding actions that might not be immediately possible. They are laws “to be fleshed out over decades.” But such regulations will provide a functional model and path forward for other countries. “Europe will be the tail that wags the American dog here,” Persily said. The persistence of GDPR has also proved another important argument: Having more personal data protection won’t break the Internet. We can consume content more or less as we did before, just with a stronger sense of security and agency.

Persily 见证了数字平台与其用户和政府之间关系的变迁。2018 年，他参与创建了 Social Science One，这是 Facebook 与学术界之间开创性的合作，最终使研究人员能够研究该社交网络的内部数据。2020 年，他们发布了一个总计超过 1EB（十亿 GB）的 URL 分享数据集，其中包含 Facebook 用户分享和点击的 3800 万个链接。Persily 认为，监管往往伴随着争议。Social Science One 的成立源于剑桥分析公司（Cambridge Analytica）丑闻，这家英国咨询公司收集了数百万Facebook 未经用户同意就利用算法，并将其用于政治竞选，包括唐纳德·特朗普的竞选。人们对算法平台日益增长的不满，使得监管变得更加开放。“人们迫切希望有所作为，”佩西利说。

Persily has witnessed the changing relationships of digital platforms to their users and governments. In 2018, he helped to create Social Science One, a groundbreaking partnership between Facebook and academics that eventually allowed researchers to study the social network’s internal data. In 2020, they released a URL Shares dataset totaling over an exabyte (a billion gigabytes), containing thirty-eight million links that were shared and clicked by Facebook users. To hear Persily tell it, regulation often follows controversy. Social Science One was incited by the Cambridge Analytica scandal, in which the British consulting firm gathered the data of millions of Facebook users without their consent and leveraged it in political campaigns, including Donald Trump’s. The mounting dissatisfaction with algorithmic platforms has created more openness toward regulation. “People desperately want something to get done,” Persily said.

在现有的策略中——包括强制算法透明度、改革第230条、规范特定类型内容的放大以及保护数据权利——透明度仍然是Persily平台监管的北极星。它之所以至关重要，部分原因是，在大多数情况下，当公司仅自行报告时，我们根本不知道信息流中到底发生了什么。透明的数据（至少研究人员可以在安全、匿名的环境中访问这些数据）是规划未来最佳监管形式的关键，但它也可能带来额外的好处，即在无需制定更明确的政策的情况下，对科技巨头施加真正的日常压力。“这将影响公司的行为。一旦他们知道自己受到监视，他们的决策就会有所不同，”Persily说道。

Out of the available strategies—ranging across enforcing algorithmic transparency, reforming Section 230, regulating the amplification of particular kinds of content, and protecting data rights—transparency remains the north star of platform regulation for Persily. It’s the most important in part because in most cases, we simply don’t know what’s really going on in our feeds when companies only self-report. Transparent data, which would be accessible at least for researchers to study in secured, anonymized environments, is key to planning the best forms of future regulation, but it may also have the added benefit of putting real day-to-day pressure on tech giants without having to craft more explicit policy. “It will affect the behavior of firms. Once they know they are being watched, then their decisions will be different,” Persily said.

让Facebook与Social Science One合作始终困难重重——该公司的律师一直辩称，与外部分析师共享任何数据都是侵犯用户隐私。这场冲突促使Persily开始构思自己关于社交媒体的联邦立法构想。他认为，未能共享符合公众利益的数据必须受到法律惩罚。“我们必须通过一些平台来了解当今的人类行为，因为大多数人类行为都发生在网上，”他说道。Persily开始起草立法措辞，并从社会科学家和政策律师那里获得反馈。Frances Haugen在Facebook上的举报让他怒火中烧，并在她到参议院作证的当天公布了草案。

It was always a struggle to get Facebook to cooperate with Social Science One—the company’s lawyers hid behind the argument that they were violating user privacy by sharing any data with outside analysts. The conflict led Persily to start working on his own concept for federal legislation addressing social media. There needed to be a legal penalty for failing to share data that was in the public interest. “It is absolutely critical that we get some window into platforms to understand human behavior today, because most human behavior is occurring online,” he said. Persily began drafting legislative language, getting feedback from social scientists and policy lawyers. Frances Haugen’s Facebook whistleblowing lit a fire under him, and he released the draft the day she testified in front of the Senate.

特拉华州参议员克里斯·库恩斯（Chris Coons）迅速联系了佩尔西利，并采纳了他的方案作为一项透明度法案的基础，随后一个更大的团体开始推动这项法案。他的构想最终成为了《平台问责与透明度法案》，该法案于2021年12月在两党支持下正式宣布。这些学术理念已逐渐成为法律。PATA要求社交媒体公司通过经审查的请求向研究人员提供数据。美国国家科学基金会。不遵守规定的公司将失去第 230 条的保护，并对其平台上的所有内容承担责任。

Chris Coons, a Delaware senator, quickly reached out to Persily and adopted his text as the basis of a transparency bill, which a larger group began pushing forward. His concept became the Platform Accountability and Transparency Act, which was announced in December 2021 with bipartisan support. The academic ideas had moved toward becoming law. PATA forces social media companies to supply data to researchers, via requests that are vetted by the National Science Foundation. Companies that don’t comply would then lose the protections of Section 230 and be liable for everything on their platforms.

PATA 加入了其他可能的立法，例如参议员艾米·克洛布查尔和辛西娅·卢米斯于 2022 年提出的两党提案《NUDGE 法案》。（该法案的缩写含义复杂，为“推动用户推动良好社交媒体体验法案”。）在美国，这些努力更像是一个开始，而非结束：我们或许已经意识到算法推送正在危害我们，无论是个人还是集体，但我们远未理解如何在政府层面最好地减轻其影响。

PATA joins other possible legislation like the NUDGE Act, a 2022 bipartisan proposal by Senator Amy Klobuchar and Cynthia Lummis. (The elaborate acronym stands for “Nudging Users to Drive Good Experiences on Social Media Act.”) In the United States, these efforts are more of a beginning than an end: we may have realized that algorithmic feeds are harming us, as individuals as well as collectively, but we’re far from understanding how best to mitigate their effects at a governmental level.

监管为那些通常被视为政治问题的问题提供了政治解决方案。诚然，算法信息流的一些最显著的问题触及政治层面，例如言论自由、骚扰、技术编码偏见以及工业资本主义等问题。但在“过滤世界”（Filterworld），信息流也影响着生活中更日常的方面。任何监管网络仇恨言论的尝试最终都会影响我们选择观看哪部电视节目或听哪张专辑的方式。算法对我们注意力的垄断也可以被打破。

Regulation offers political solutions for what is often observed as a political problem. It’s true that some of the most visible problems of algorithmic feeds broach the political, with issues of free speech, harassment, technologically encoded bias, and industrial capitalism. But in Filterworld, the feeds impact more mundane aspects of life, too. Any attempt at regulating online hate speech will also end up influencing the way we choose which television show to watch or which album to listen to. Algorithms’ monopoly on our attention can be broken down, too.

然而，在文化问题上，监管并非唯一的答案。（政府政策在这方面很少奏效。）法律可以强制平台封禁有问题的内容，但它无法迫使Spotify推荐更具挑战性或更具创意的音乐播放列表。不幸的是，我们没有宪法赋予的个人品味权利。因此，我们也必须改变自身的习惯，更加清晰地了解我们如何消费文化，以及如何抵制算法推送的被动路径。就像我们可能会选择在杂货店购买有机食品一样，我们必须寻找支持非同质化文化的数字空间，让艺术家能够在不受千篇一律的压力下表达自我。

Yet regulation cannot be the only answer when it comes to culture. (Governmental policy rarely succeeds in that arena.) A law can force a platform to ban problematic content, but it can’t make Spotify recommend you a more challenging or creatively interesting playlist of music. Unfortunately, we do not have a constitutional right to personal taste. Therefore, we also must change our own habits, becoming more aware of how we consume culture and how we can resist the passive pathways of algorithmic feeds. The same way we might choose to buy organic-labeled food in a grocery store, we have to seek out the digital spaces that support nonhomogenized culture and that allow artists to express themselves without the crushing pressure toward sameness.

我们必须谨慎选择关注的动态，并准确理解我们的注意力是如何转化为对创作者的经济支持。定向广告可能是最糟糕的选择之一。数字世界就像一片森林；Facebook 和 TikTok 可能是参天大树，遮挡了大部分的视线。阳光，但如果你去寻找，就会发现在它们的阴影下还存在着其他的可能性。

We have to be careful about which feeds we follow as well as understand precisely how our attention is transformed into economic support for creators. Targeted advertising might be one of the worst possibilities. The digital landscape is like a forest; Facebook and TikTok might be towering trees, blocking out much of the sunlight, but there are other possibilities growing in their shadows if you look for them.

目前已经有办法直接向艺术家支付他们在线创作的费用。Bandcamp 相当于独立音乐人的数字唱片商店；用户可以直接购买数字文件和流媒体访问，而无需通过 Spotify 进行交易。Patreon 允许创作者为任何他们想看的内容设置付费墙，无论是文字、图片还是音频。它提供线性的帖子推送，用户可以付费订阅，这种关系比 Twitter 的关注要强大得多。Substack 也为电子邮件简报订阅提供了同样的功能。

There are already ways to pay artists directly for what they create online. Bandcamp serves as the digital equivalent of an indie record store for independent musicians; users can buy digital files and streaming access directly instead of Spotify mediating the exchange. Patreon enables creators to paywall anything they choose to, whether writing, images, or audio. It offers a linear feed of posts that the user subscribes to monetarily, a much more powerful relationship than a Twitter follow. Substack does the same for email newsletter subscriptions.

2008年，《连线》杂志编辑凯文·凯利（Kevin Kelly）曾写道，创作者只需找到“1000名真正的粉丝”来资助他们的作品并维持生计——这一千人每年可能会付给他们100美元。这与大型数字平台的模式截然不同，大型数字平台的受众必须尽可能多。“争取一千名客户比争取一百万名粉丝要现实得多，”凯利写道。然而，所有这些小型平台在努力发展壮大并服务于最大数量的用户、创作者和消费者时，都面临着变得更加算法化的诱惑。它们也不断受到大型科技公司的威胁，这些公司可能会收购它们或将其淘汰。没有人能保证它们的非算法功能会保持不变。

In 2008, the Wired editor Kevin Kelly famously wrote that a creator needs to find only “1,000 true fans” to fund their work and allow them to make a living—one thousand people who might pay them $100 a year. It’s an entirely different model from the larger digital platforms, where audiences have to be as big as possible. “A thousand customers is a whole lot more feasible to aim for than a million fans,” Kelly wrote. Yet all of these smaller platforms face the temptation of becoming more algorithmic as they try to grow and serve the maximum number of users, creators and consumers alike. They are also constantly threatened by giant tech companies that could acquire them or stamp them out. There is no guarantee their non-algorithmic functionality will stay the same.

最有力的选择或许也是最简单的：停止关注那些利用技术的平台。有一种方法可以做到这一点，同时继续使用数字技术，坚持使用那些对用户更友善的网站和公司。我们可以回归更加自主的互联网。但更戏剧性的选择是完全退出，重新思考如何维持线下文化。过去十年的数字身份危机让改变看起来遥不可及。Facebook 像亮蓝色的数字葛藤一样，慢慢地扼杀了开放的互联网。随着这个平台变得不可避免，我感觉自己也对它失去了兴趣，因为它不可能同时满足所有目的并在所有方面都取得成功。如果说 Facebook 是一切，那么它也什么都不是，是一个毫无差别的内容堆。其他平台也以同样的方式衰落：马克·扎克伯格领导下的 Instagram，丹尼尔·埃克领导下的 Spotify，埃隆·马斯克领导下的 Twitter。到 2023 年，社交媒体似乎已经进入一个新阶段，其缺点比优点更加明显。

The most powerful choice might be the simplest one: Stop lending your attention to platforms that exploit it. There is a way to do that while continuing to use digital technology, sticking to websites and companies that treat users better. We can return to a more DIY Internet. But the more dramatic option is to log out entirely and figure out how to sustain culture offline once more. The digital identity crisis of the past decade has made change seem impossible. Facebook slowly choked the open Internet like bright-blue digital kudzu. As the platform became inescapable, I felt myself losing interest in it as well, because it’s impossible to serve every purpose at once and remain successful at all of them. If Facebook was everything, it was also nothing, an undifferentiated mass of content. Other platforms decayed in the same way: Instagram under Mark Zuckerberg, Spotify under Daniel Ek, Twitter under Elon Musk. By 2023, social media seems to have entered a new phase in which its downsides are more obvious than its advantages.

如今，我在网上感到有些压抑，部分原因是我无法像以前那样自由地表达自己，不像以前那样能用个人博客和慢慢发展起来的与他人的对话了。模板太过局限，节奏也太快了。尽管二十年前的技术比现在差得多，但体验，或者说生态系统，自有其优势。我认为有办法重新找回一些创造力，那种低保真可能性和自由的氛围。虽然监管可以对算法推送提供最低限度的控制，但文化的重建是一个不同的过程，更像是种植和培育花园。这需要时间。首先，我们需要找到合适的数字结构，然后我们需要每天努力探索新的网络生活方式。

I feel somewhat stifled online these days, in part because I can’t express myself as much as I once could, in the days of personal blogs and slowly developed conversation with other people. The templates are too restrictive, and the pace is too fast. Even though the technology was much worse two decades ago, the experience, or the ecosystem, had its advantages. I think there’s a way of recapturing some of that creative energy, the atmosphere of lo-fi possibility and freedom. While regulations can provide for some minimum of control over algorithmic feeds, the reconstruction of culture is a different process, more like planting and cultivating a garden. It takes time. First, we need to seek out the appropriate digital structures, and then we need to carry out the daily labor of determining a new way of living online.

我的算法清理

MY ALGORITHM CLEANSE

到了2022年夏天，我完全被困在了“过滤世界”（Filterworld）里。两年的疫情让我的生活大部分时间都依赖于数字信息流：与朋友互动；通过流媒体服务观看电视和电影；以及通过推特获取周围世界的实时新闻（而且从来就没有新闻）。所有接触到的媒体和文字信息都是通过数字平台传递的，而这些过程我几乎无法掌控。我的手机就像粘在手上一样，成为我消磨每一秒无刺激闲暇时光的工具。这些信息流确保我一天24小时都能看到新鲜事物，无论我关注的人最近是否发布了新内容。TikTok 是这种“永久信息流”的终极实现。无论我是凌晨三点辗转反侧，还是在午后遛狗，抑或是在晚餐时间躲进餐厅洗手间，我都能随时访问到那些令人耳目一新的新内容流。

By the summer of 2022, I had become wholly trapped in Filterworld. Two years of the pandemic had made me dependent on digital feeds to conduct much of my life: interacting with friends; watching television and movies through streaming services; and getting real-time news about the world around me through Twitter (and there was never not news). All the media and text that reached me was mediated through digital platforms in processes I had little control over. My phone was glued to my hand as a tool to soak up any spare second of nonstimulation. The feeds ensured that I would always have something new to look at, twenty-four hours a day, no matter whether the people I followed had posted anything recently or not. TikTok was the ultimate fulfillment of the forever-feed. It didn’t matter whether I was awake and restless at three a.m., walking our dog in the middle of the afternoon, or in a restaurant bathroom during dinner, I could always access new content streams that were always refreshing.

这些信息流渗透到我的生活是多么顺畅，怎么说都不为过。关注这些信息流就像整天不停地抽烟，一次又一次地狂扫信息，从起床刷推特了解隔夜新闻，到深夜浏览Netflix主页决定看什么。我对这种情况的不满由来已久，疫情更是雪上加霜；作为一名作家和各种文化的消费者，这些平台是我与那些让我接触新鲜有趣事物的人建立联系的地方。我很感激多年来在Twitter和Instagram上投入的时间，以及在这些平台上建立的个人和职业关系。但我开始思考，尽管这些动态信息带给我一些我原本不可能看到或听到的东西，但我对它们的过度依赖也让我与过去十年中我早已遗忘的另一个体验领域隔绝：与稀缺而非无限的相遇，与自己在特定时刻想要看到什么的判断和选择过程，而无法滚动屏幕离开。

It’s hard to overstate just how smoothly the feeds infiltrated my life. Following them all was like chain-smoking throughout the day, one information binge at a time, from the wake-up Twitter scroll to find out about overnight news to the nighttime crawl of the Netflix home page to determine what to watch. My dissatisfaction with this situation was a long time coming, exacerbated by the pandemic; as a writer and a consumer of all kinds of culture, these platforms were the place where I connected with the people who exposed me to new and interesting things. I appreciated the years that I had invested in being on Twitter and Instagram, the personal and professional relationships I had developed in those spaces. But I began to think that as much as the feeds brought me things I never would have seen or heard otherwise, my overdependence on them was also cutting me off from a different realm of experiences that I had forgotten about over the course of the decade: the encounter with scarcity rather than infinity, the process of judging and choosing for myself what I wanted to see in a given moment, without the option to scroll away.

如果说算法焦虑的一种形式是感觉被算法推荐误解，那么另一种形式则是感觉被算法推荐劫持，感觉即使想摆脱也逃脱不了。或许现在太多东西都依赖于这些信息流，它们的影响也过于普遍。迄今为止，监管力度不足，而这种体验又如此引人入胜、便捷，以至于大多数互联网用户都难以放弃。算法推荐之所以令人上瘾，是因为它们总是在潜移默化地印证你自身的文化、政治和社会偏见，将你周围的环境扭曲成你自己的镜像，同时也对其他人做着同样的事。这让我感到焦虑，我对自己通过互联网生活的看法，很可能只是由信息流塑造的虚构。我对朋友在某一天的活动、各个城市发生的事情、哪些新闻重要，甚至天气的感知，很大程度上都取决于我在自动化应用程序上看到的内容。更重要的是，这些信息流越来越支离破碎，漏洞百出，几天前的帖子就像刚发生过一样呈现。最终，我的自我意识被那些隐形受众的反馈所左右，而他们的注意力也受到了算法的操控。我不确定如果没有算法推荐，我会变成什么样；我不知道其他在数字平台上度过了多年时光的人是否能完全确定这一点。一种恐惧油然而生：在被动消费我感兴趣的东西时，我是否放弃了自主权，去探索什么对我来说真正有意义？

If one form of algorithmic anxiety is about feeling misunderstood by algorithmic recommendations, another is feeling hijacked by them, feeling like you couldn’t escape them if you tried. Perhaps too much now depends on these feeds, and their influence is too pervasive. Regulation, so far, has been less than effective, and the experience so compelling and convenient that most Internet users struggle to give it up. Algorithmic recommendations are addictive because they are always subtly confirming your own cultural, political, and social biases, warping your surroundings into a mirror image of yourself while doing the same for everyone else. This had made me anxious, the possibility that my view of my own life—lived through the Internet—was a fiction formed by the feeds. So much of my perception of what my friends were up to on a given day, what was going on in various cities, which news stories mattered, even the weather, was dictated by what I saw on automated apps. What’s more, those feeds were all increasingly fractured and flawed, presenting posts from days ago as if they had just happened. Ultimately, my sense of self was beholden to the responses I got from my invisible audiences, whose attention was algorithmically mediated, too. I wasn’t sure who I would be without algorithmic recommendations; I don’t know that anyone else who has spent years of their life on digital platforms can be totally sure. A fear took hold: In passively consuming what I was interested in, had I given up my agency to figure out what was truly meaningful to me?

为了对抗这种即将来临的徒劳，我做了唯一能做的事情。我决定试试看，没有这些推送，我的生活还能过得下去吗？就像有些人会在四旬斋期间戒糖，或者在一月份戒酒一样，我会暂时不去想他们。我会拒绝他们提出的帮我思考的诱人提议，尝试自己去做。

To combat that looming futility, I did the only thing I could do. I decided to see if I could live my life without these feeds, go without them for a while the way some people might give up sugar for Lent or alcohol in January. I would say no to their attractive offer of thinking for me and try to do it myself.

这听起来很容易做到。我只需要从手机上删除一些应用程序，并退出一些我在笔记本电脑上经常访问的网站的账户。但与此同时，这感觉又不可能实现。我的工作生活离不开推特，我的社交生活离不开Instagram，我的音乐生活离不开Spotify。我担心自己会错过一些重要的事情，比如远距离观察朋友的生活动态，阅读一篇新的喜欢的文章，或者抓住一个只有在网上才能找到的工作机会。“害怕错过”（FOMO）这个词太轻描淡写了——尽管所有证据都表明并非如此，但我担心，如果我不再像以前那样频繁地使用社交媒体，不再持续关注社区动态，那我就等于没有存在感了。如果算法没有记录我，那我还算参与到这场浩瀚的社区对话中吗？在网上发布信息就像艺术家河原温 (On Kawara) 从 1969 年开始创作的系列作品“我还活着”，在这个系列中，他发出了数百封带有标题短语的电报，似乎在不断证明这一点。

It sounded like an easy enough thing to accomplish. All I had to do was delete some apps from my phone and sign out of some accounts on sites that I drifted toward far too easily on my laptop. But at the same time, it felt impossible. My work life was on Twitter, my social life was on Instagram, my music was on Spotify. I worried I would miss out on something vital, like observing a friend’s life event from afar, reading a new favorite essay, or grabbing a work opportunity that would have come to me only online. FOMO was too light of a phrase—I feared, against all evidence to the contrary, that not to be on social media in the constant way I had been, always participating in the communal feed, would be not to exist. If the algorithm wasn’t registering me, then was I even participating in this vast communal conversation? Posting online is like the artist On Kawara’s series “I Am Still Alive,” begun in 1969, in which he sent out hundreds of telegrams with the titular phrase, as if to constantly prove it.

然而，当我仔细思考自己的焦虑时，我意识到它们其实无关紧要。如果我看不到朋友度假的十几张照片，或者看不到那些被大肆炒作的小说的最新评论，或者看不到推特上某个特定时刻哪些热门话题在热议，我会错过什么呢？在我日复一日的现实生活中，这些内容几乎没有任何影响。我害怕失去某种联系，但这种联系毕竟比遛狗时和邻居聊天更自然、更不直接。我想，在我一直上网、追逐算法推送信息时进入的持续高度警觉状态和完全无知的状态之间，一定存在一个更平衡的状态。我到底需要多少数字输入？对脱离的恐惧最终可能比现实更糟糕。

Yet when I thought twice about my anxieties, I realized they were relatively inconsequential. What would I really miss if I didn’t see a dozen photos of a friend’s vacation, or the latest reviews of hyped-up novels, or which viral arguments were dominating Twitter at a particular moment? In the context of my day-to-day, physical life, these bits of content had almost no impact. I feared the loss of some connection, but that connection is, after all, more ambient and less direct than chatting with neighbors when I walk my dog. I figured that there must be a happier medium between the state of constant hyperawareness I entered while being online all the time, chasing algorithmic feed updates, and a state of total ignorance. How much digital input did I really need? The fear of disengaging might turn out to be worse than the reality.

我开始把我的实验想象成一场算法净化，一场信息节食，理想情况下，它能让我比以前更健康。我还想知道，在没有那么多推荐的情况下，如何才能跟上当代文化。但我一直拖延着这个承诺。终于，在一个周末，2022年8月，跟上Twitter信息流的步伐变得比退出更令人疲惫。当时，喧闹的书呆子之王埃隆·马斯克（Elon Musk）凭借收购一家电动汽车公司成为全球首富之一）威胁要直接收购Twitter，将其变成自己的私人游乐场（他最终在当年10月成功了）。Twitter平台上的情绪变得比以往更加低落。Instagram也变得越来越难以使用，因为它的信息流开始优先推广视频片段，引发了用户的强烈不满，试图追赶TikTok。

I began thinking of my experiment as an algorithm cleanse, an information diet that would ideally leave me healthier than before. I also wanted to find out what it would take to follow contemporary culture without the assistance of so many recommendations. But I kept putting off the commitment. Finally, one weekend in August 2022, the prospect of keeping up with the feeds became more exhausting than exiting from them. At the time, the boisterous nerd king Elon Musk, one of the wealthiest men in the world thanks to his takeover of an electric car company, was threatening to buy Twitter outright and turn it into his private playground (he finally succeeded that October). The mood on the platform had soured, turning even more negative than usual. Instagram had become increasingly unusable, too, as its feed began promoting video clips above all—provoking mass complaint from its users—in a race to catch up with TikTok.

TikTok本身就是一种麻木，因为它对用户输入的响应就像一种读心术，让我完全不需要思考。我可以暂时逃离现实五分钟（或者更可能是十五分钟），几乎不会受到任何干扰，就像一次鼠尾草之旅，让你短暂地完全脱离现实世界，尽管感觉更愉悦。TikTok的“为你推荐”动态实现了大卫·福斯特·华莱士的小说《无尽的玩笑》中“娱乐”的理念：内容如此引人入胜，让人欲罢不能。尽管它很容易让人上瘾，但华莱士将他虚构的“娱乐”描述为“空洞无物，缺乏戏剧性的走向——没有叙事走向真实故事”——这也恰如其分地描述了TikTok正在向无形的氛围和感受转变，远离连贯的信息。沉迷于这些科技产品当然不会让我变得更聪明，也不会让我产生更复杂的想法，而这大概是我作为一名作家的职责所在。

TikTok itself was a kind of numbing, since its responsiveness to user input felt like a form of mind-reading, removing my need to think at all. I could take five-minute (or more likely fifteen-minute) breaks from immediate reality with minimal disruption, like a salvia trip that takes you utterly and briefly out of the world, albeit more pleasant. The TikTok “For You” feed was a fulfillment of “the Entertainment” in David Foster Wallace’s novel Infinite Jest: a piece of content so compelling that no one could stop watching. As addictive as it was, Wallace described his fictional Entertainment as “oddly hollow, empty, no sense of dramatic towardness—no narrative movement toward a real story”—which is an apt description, too, of TikTok’s drift toward formless vibes and feelings, away from coherent information. Surely bingeing on these technologies wasn’t making me any smarter or able to generate more complex thoughts, which was presumably my job as a writer.

我离开社交媒体时，没有在各种动态上说再见或宣布。公开承认清理账号无异于雪上加霜，更何况这还是社交媒体鼓励的那种自我吹捧。没人会在意你什么时候停止发推文；算法只会把更愿意参与的内容塞进去，因为在过滤世界里，每个人都是可以替代的。你的大多数粉丝不太可能注意到你的离开，因为至少在算法看来，你这个沉寂的账号已经不再值得优先关注了。

When I left social media, I didn’t say goodbye or make an announcement on my various feeds. To publicly acknowledge the cleanse was to jinx it, not to mention an example of the kind of self-aggrandizement that social media encourages. No one cares when you stop tweeting; the algorithm will simply slot in the content of some more willing participant, because, in Filterworld, everyone is replaceable. The majority of your followers are unlikely to even notice your absence because, at least to the algorithm, your sleepy account is deemed no longer worthy of prioritization.

于是，在八月的一个周五晚上，我干脆在下班后就断网了。当然，我之前已经退出过很多次了。但我知道我暂时不会再登录了——我计划了几个月的“净化”——沉默震耳欲聋。社交媒体一直是我全天候获取无限实时信息的门户，而我的手机却突然变成了一块静电砖。

So, on a Friday evening in August, I simply cut myself off at the end of the workday. Of course, I had logged out plenty of times before. But knowing that I wasn’t going to get back on for a while—I planned for a few months of cleanse—the silence was deafening. While social media had been a 24/7 portal to infinite real-time information, my phone suddenly became a static brick.

第一个周末还好。周末本来就应该下线，不过，我得工作几个小时，所以很怀念推特上那种轻松愉快的氛围，就像教室里的孩子都困在教室里休息一样。然而，接下来的星期一却像个折磨。我的拇指渴望滑动屏幕，大脑也因为没有了持续不断的信息冲击而出现戒断反应。这些症状类似于焦虑的生理表现：神经抽搐、脾气暴躁、浑身不适。这种变化非但没有让我放松，反而让我对信息的缺失感到不安。或许我有点夸张，但变化也确实很大。我每天浏览的信息和多媒体从数百甚至数千条变成了寥寥无几。通过切断算法推送，我彻底摆脱了数字消费习惯。

The first weekend wasn’t so bad. Weekends are when we’re supposed to be logged out anyway, though when I had to do a few hours of work I missed the ambient camaraderie of Twitter chatter, like a classroom of kids all stuck inside at recess. The following Monday, however, was torture. My thumb itched for swiping, and my brain went through withdrawals from not having the constant onslaught of information. The symptoms were akin to the physical manifestations of anxiety: nervous twitching, short temper, general discomfort. Far from being relaxed by the change, I was perturbed by the absence. Perhaps I was being dramatic, but the difference was dramatic, too. I had gone from seeing hundreds, or even thousands, of individual bits of information and multimedia a day to just a handful. By cutting out algorithmic feeds, I had stomped on the brake pedal of my digital consumption habits.

为了纯粹的刺激，我在手机上下载了几个“坐立不安”应用——这些游戏可以让你堆积木或拨动电灯开关，相当于电子版的“焦虑珠”。其中一个名为“抗压力”的应用提供了一些选项，比如用吸尘器把脏兮兮的地板吸干净。它们缓解了我的烦躁，但却像对着虚空发推文一样毫无意义——毫无意义的行为。于是，我在笔记本电脑上创建了一个文件，名为“推文，未发”，把所有我本来会发布的观察记录都写在里面。回头看看这份文件，那些草稿毫无启发性：“马克思主义者诺拉·艾芙隆：‘一切都是资本。’”“我怀念用翻盖手机发短信244244就能发推文的时代。”这些玩笑或许能获得一些点赞，但在推特之外却毫无意义。毕竟，我的感知力已经进化到可以适应推特的地步了。

For pure stimulation, I downloaded a few “fidget apps” onto my phone—games that allow you to stack blocks or flip light switches, a digital iteration of worry beads. One called Antistress offered options like vacuuming a dirty floor into satisfying cleanness. They assuaged my restlessness but were just as pointless as tweeting into the void—action without meaning. Instead, I started a file on my laptop titled “Tweets, Not Tweeted,” where I wrote down every observation I would have otherwise posted. Looking back at the document, the drafts are not inspiring: “Marxist Nora Ephron: ‘Everything is capital.’ ” “I miss the days when you could tweet by texting 244244 on your flip phone.” These were jokes that might net a handful of likes, but they made no sense outside of Twitter itself. After all, that was the venue my sensibility had evolved to fit.

就像任何戒断反应一样，净化让我变得暴躁。不再能从那些用点赞和转发来鼓掌的隐形观众那里获得持续的反馈，我在现实生活中对杰西说了太多平凡的言论，而当她（理所当然地）忽略了大部分内容时，我就会很恼火。我不断地坐立不安，四处寻找刺激，这让她分心不已，以至于杰西经常在深夜逃离公寓，去遛狗，时间异常长。

As with any kind of withdrawal, the cleanse made me cranky. No longer getting constant feedback from an invisible audience applauding with faves and retweets, I lobbed too many mundane observations at Jess in real life and got annoyed when she (deservedly) ignored most of them. My constant fidgeting and casting about for stimulation became so distracting that Jess often fled the apartment for unusually late-night, unusually long dog walks.

在接下来的几周里，我发现互联网的设计已经脱离了算法信息流的束缚。这些信息流就像重要的航道，确保各种形式的内容都能到达目标受众，而无需创作者和消费者付出太多努力。创作者，无论是个人、品牌还是出版物，都可以发布内容，并相信内容会触达关注者，尽管效率并非100%。消费者可以打开信息流，看到吸引他们注意的内容，至少大多数时候是这样。博客和其他旨在汇总头条新闻和突出趋势的网站，例如最初的Gawker.com，在信息流的冲击下消失了。由于这个系统根深蒂固，出版物网站纷纷拆分主页，以至于它们通常一次只显示几篇报道，图片最多，文字最少。当我浏览这些网站时，我感觉自己像个不速之客，一个不该出现在那里的人。这些网站几乎都在喊：难道你不知道你应该使用 Facebook 或 Twitter 吗？

In the following weeks, I discovered that the Internet wasn’t really designed to function without algorithmic feeds anymore. These feeds worked like vital shipping lanes, ensuring various forms of content reached their intended targets without either the creator or the consumer having to work very hard. The creator, whether a person, brand, or publication, could post something and trust that it would reach people who were following them, though not with 100 percent efficiency. The consumer could open their feed and see content that was compelling to them, at least most of the time. Blogs and other websites whose function was to aggregate headlines and highlight trends, like the original Gawker.com, had vanished in the onslaught of feeds. Because this system was so entrenched, publication websites had dismantled their home pages to the point that they often featured only a few stories on the screen at a time, with a maximum of images and a minimum of text. When I browsed them, I felt like an unexpected visitor, someone who wasn’t supposed to be there. The sites all but shouted: Don’t you know you’re supposed to be on Facebook or Twitter!?

在清理过程中，我还发现推荐系统会在意想不到的地方出现。我最终转向使用《纽约时报》 App作为查看新闻的主要方式，但这款App有一个“为你推荐”标签，很像TikTok的，它会根据你之前的操作推荐一系列你认为可能会点击的文章。它立刻把我归类到艺术和文化类文章中，偶尔还会推荐一些房地产色情内容，我很喜欢这一点。但这种有限的视野正是我想要避免的。我停止使用这个标签，却发现很难在App的其他平台上获得全面的新闻视野。屏幕上仍然只显示少数编辑认为重要但并不总是符合我兴趣的内容。我不得不面对自己对个性化的期望。

During my cleanse, I also discovered that recommender systems pop up in unexpected places. I eventually turned to the New York Times app as my primary way of checking in on news, but that app features a “For You” tab, much like TikTok’s, that uses your previous actions to suggest a slew of articles you are deemed likely to click on. It immediately pigeonholed me into arts and culture pieces, with the odd item of real-estate porn, which I appreciated. But that kind of limited horizon was exactly what I was trying to avoid. I stopped using that tab, yet found it difficult to get a broad view of the news elsewhere on the app. The screen was still limited to a handful of items that the editors deemed important but that didn’t always line up with my interests. I had to confront my own expectation of personalization.

2022年9月，女王伊丽莎白二世去世那天，我已经很久没看新闻了，杰西几个小时后才通知我。后来，我无聊地用《纽约时报》的App读了大卫·布鲁克斯的一篇专栏文章——这种写作方式——以及他是一位我以前从未感兴趣的作家——她考虑过进行干预。我突然明白了那些老情景喜剧里，有人屈服的场景。早餐桌上的一份报纸，引用了一个平庸的标题。没有了小众的推特闹剧，这才是新鲜感的主要来源。

The day Queen Elizabeth II died in September 2022, I hadn’t checked on the news in a while, and Jess only informed me hours later. When I ended up, out of boredom, using the NYT app to read an op-ed by David Brooks, a form of writing—and a writer—I’d had no previous interest in, she considered staging an intervention. I suddenly understood those old sitcom scenes when someone folds a newspaper at the breakfast table and cites a banal headline. Bereft of niche Twitter drama, this was the main source of novelty.

完全摆脱算法几乎是不可能的。毕竟，谷歌搜索仍然受算法驱动，每个电子邮件客户端都会在一定程度上自动对邮件进行排序——我没法完全关闭垃圾邮件过滤器。由于我几乎完全没有地理方向感，如果没有谷歌地图推荐的路线，我会迷路。但我仍然可以放弃那些我曾经依赖的文化信息流。在没有推荐的情况下，我只能选择自己主动选择阅读的内容，比如电子邮件简报。这些邮件相当于手印小册子的数字版，提供了一种直接联系我想听到的出版物或作家的方式——我信任的声音。简报格式再次流行起来，正是为了避免算法信息流的主导影响。我欣赏它们，因为它们就像印刷杂志一样，内容是有限的，与信息流截然相反。

Escaping algorithms entirely is nearly impossible. Google Search is still driven by an algorithm, after all, and every email client automatically sorts messages to some degree—I couldn’t very well turn off my spam filter. Since I almost entirely lack a sense of geographical direction, I’d be lost without the Google Maps recommended routes. But I could still go without those primary feeds that I had relied on for cultural consumption. In the absence of recommendations, I was left with things that I intentionally chose to consume, like email newsletters. The digital equivalent of hand-printed pamphlets, these missives offered a way to connect directly with publications or writers who I wanted to hear from—voices that I trusted. The newsletter format had become popular once more precisely to avoid the dominant influence of the algorithmic feed. I appreciated them because, like a print magazine, they were composed and finite—the opposite of a feed.

正如我所希望的那样，我开始更频繁地一次性阅读长篇文章，也减少了浏览器打开的标签数量，因为我不再面临一大堆其他选项。一个月后，我那如痴如醉的推特写作欲望消退了，或者至少转向了更长的日记式写作，不再局限于280个字和精辟的笑话。在我的未发推文列表中，我反而发现，发布在推特上的想法与正常、连贯的想法截然不同，因为它们被剥离了背景，被迫以环境意识碎片的形式存在。（俗话说，“推特不是现实生活。”）但最引人注目的是我与摄影关系的转变。我的iPhone相机没有任何变化，我仍然随身携带，但失去了Instagram这个公开平台，我的拍照欲望就少了很多。我拍的几张照片也有所不同：它们是我想为自己捕捉的照片，通常比Instagram的既定审美更怪异或更丑陋。我在派对和晚宴上拍的照片少了，更多是拍摄城市街景和附近亮着灯的狗狗公园，这些照片在应用程序的框架下看起来不太好看。（不过，我狗狗的快照并没有减少。）

As I had hoped, I began reading long articles in a single sitting more often and left fewer tabs open in my browser, since I wasn’t faced with a cascade of alternative options. The graphomanic desire to tweet subsided after a month, or at least turned toward longer diaristic writing, not limited to 280 characters and pithy jokes. In my list of untweeted tweets I instead observed how little the kind of thoughts that get posted on Twitter resemble normal, coherent thoughts, as they are shorn of context and forced to exist as atomized bits of ambient consciousness. (As the saying goes, “Twitter is not real life.”) But it was the shift in my relationship to photography that was the most striking. Nothing had changed about my iPhone camera, which I still had in my pocket at all times, but without the public venue of Instagram I had much less desire to take photos. The few photos I did take were different, too: they were images that I wanted to capture for myself, more often weirder or uglier than the established aesthetic of Instagram. I took fewer photos at parties and dinners and more of city streetscapes and my neighborhood dog park lit up in the dark, images that wouldn’t have read well in the app’s frame. (Snapshots of my dog did not decrease, however.)

从表面上看，我的生活可能并没有因为虽然算法清理让我感觉思路清晰，心绪也变得不那么拥挤了。我意识到，意向性是赋予事物意义的一部分：我之所以能够更加欣赏每个故事、照片或相册，是因为我主动去寻找它们。这也意味着我必须更加努力地去寻找我想要的东西，放弃那些自动化内容高速公路的帮助。实验进行到第二个月，当我调整了习惯后，我开始感到一种怀旧之情。这让我想起了青少年时期，在主流社交媒体出现之前，我是如何与互联网互动的。

Outwardly, my life may not have been dramatically altered by the algorithm cleanse, but I did gain a sense of clarity and a less crowded mental landscape. I realized that intentionality was part of what gave things significance: I could appreciate each story, photo, or album more because I had chosen to seek it out. That also meant I had to work harder to find what I wanted, forgoing the help of those automated content highways. By the second month of my experiment, when I had adjusted my habits, I began to feel a sense of nostalgia. It reminded me of how I interacted with the Internet as a teenager, back before mainstream social media existed.

算法之前的数字文化

DIGITAL CULTURE BEFORE ALGORITHMS

1998年，日本艺术家兼作家阿部芳年推出了一部名为《连续实验玲音》（Lain）的动画电视作品。几年后，由于早期互联网的兴起，十几岁的我在美国发现了这部作品。它不仅塑造了我的审美观念，也让我开始思考网络生活的意义。阿部芳年的作品在美国并不出名，尤其是与吉卜力工作室的现象级作品相比，但它融合了卡夫卡小说的文学黑暗与梵高绘画的视觉冲击。《玲音》（Lain）通常被简称为《玲音》，是一部关于互联网时代生活的寓言故事。在这部作品中，一个名叫玲音的少女发现了一个名为“连线”（Wired）的虚拟世界。作品的色调幽暗而柔和，营造出深夜亮灯卧室的氛围，而玲音大部分时间都待在卧室里。这让我想起了自己家里放置电脑桌面的地下室，在那里我常常发现自己虽然孤身一人，但却沉浸在互联网抽象的凝聚力之中。

In 1998, the Japanese artist and writer Yoshitoshi ABe released an anime television show called Serial Experiments Lain. When I discovered it a few years later as a teenager in the United States, thanks to the early Internet, it became a formative part of my aesthetic sensibility as well as a way to question what it meant to be online in the first place. ABe’s work isn’t very well known in the United States, particularly when compared to a phenomenon like Studio Ghibli films, but it combines the literary darkness of a Kafka story with the visual impact of a van Gogh painting. Lain, as it’s usually shortened, is a fable of life in the Internet era. In the show, a teenage girl named Lain discovers a virtual realm called “the Wired.” The show’s color palette is dark but soft, sunk into the mood of a lit bedroom late at night, which is where Lain spends much of her time. It reminded me of the basement where the computer desktop was stationed in my own home, where I often found myself physically alone but immersed in the abstracted togetherness of the Internet.

这部剧震撼人心，正如伟大的艺术作品应有的那样。在莱恩的世界里，“连线”（Wired）连接着地球上所有通讯方式：电视、电话和互联网。它们共同构成了一个横跨物理空间和虚拟空间的完全合成现实——莱恩意识到，网络上发生的一切都会影响现实世界。通过“连线”的经历，莱恩发现了自己超越腼腆初中生身份的自我。对她来说，“连线”成为了一个完全独立的空间，她可以通过它来定义自己。

The show was mind altering, as great art should be. In Lain’s world, the Wired is an interconnection of all communication on Earth: television, telephone, and the Internet. Together, they form a total synthetic reality that exists across physical and virtual space—recognizing that whatever happens online influences the real world, too. Through her experiences in the Wired, Lain discovers her own identity beyond that of a shy junior high schooler. It becomes a space of total independence for her, through which she can define herself.

尽管我对算法推送犹豫不决，但我永远不会彻底放弃互联网吧，因为它在我这一生中带给了我太多。它的好处仍然大于坏处。就像莱恩一样，互联网定义了我的人生。我追寻的问题并非我们是否应该放弃数字生活，而是如何改善它们，让它们更有价值。（全球化的世界似乎不太可能再次放弃数字网络。）从我童年时期父母在我家地下室用拨号上网连接一台老旧的台式电脑开始，互联网就为我提供了逃避现实的空间，一个探索新文化、结识新朋友、构建超越我所能直接接触的世界观的空间，就像一个亚历山大图书馆——24小时开放，总是有乐于助人的工作人员。我的家人鼓励我所有的兴趣，但却无法让我接触到我在网上找到的那些小说、音乐、电子游戏和电视节目。当我感觉自己被困在康涅狄格州郊区时，没有任何繁荣的街头生活或文化机构可以拯救一些蹩脚的音乐场所，互联网是我所能接触到的最具文化激进的空间。

Despite my hesitancy around algorithmic feeds, I could never give up the Internet entirely, because it has brought me too much over my lifetime. Its positives still outweigh its negatives. As it did for Lain, the Internet has defined my life. The question I’m pursuing is not whether we should abandon our digital lives, but how we can improve them, make them even more valuable. (A globalized world seems unlikely to ever forgo digital networks again.) From the time my parents connected dial-up to an aging desktop computer in the basement of my childhood home, the Internet has offered escape, a space to discover new bodies of culture, to meet new people, and to construct a worldview that went beyond what I had immediate access to, like a Library of Alexandria—open twenty-four hours a day, always staffed with helpful people. My family encouraged all of my interests, but couldn’t have introduced me to the kinds of novels, music, video games, and television that I found online. When I was feeling marooned in the Connecticut suburbs, without any thriving street life or cultural institutions to turn to save a few grungy music venues, the Internet was the most culturally radical space within my reach.

我想很多美国千禧一代也有同样的感受；就像拥有自己的汽车一样，开放的互联网提供了即时的自主决定权。我们每个人在培养个人品味、弄清楚自己喜欢什么、不喜欢什么方面都走着不同的道路。但我们发现自己的方式却大同小异，这都由我们所处的时代和周围的科技所决定。前几代人在青少年时期或许有舞厅或独立电台来帮助他们发现新音乐，21世纪的年轻人有TikTok动态和Spotify播放列表，而20世纪90年代末和21世纪初的千禧一代则拥有在线论坛和MP3盗版。与算法推送的顺畅渠道相比，这些都需要付出更多的努力才能找到自己喜欢的东西并消费它。虽然避免这种劳动可能很方便，但它也让我们的个人品味变得更加脆弱，更难获得。

I imagine much of the American millennial generation feels the same way; like getting your own car, the open Internet provided an immediate freedom for self-determination. We all take different paths in developing our sense of personal taste, figuring out what we like and what we don’t. But the modes that we discover ourselves through are similar, dictated by our era and the technology that surrounds us. While previous generations might have had dance halls or independent radio stations to help them discover new music during their formative teenage years, and young people in the twenty-first century have TikTok feeds and Spotify playlists, millennials in the late 1990s and early 2000s had online forums and MP3 piracy. These required much more labor to find what you like and consume it than the frictionless avenues of algorithmic feeds. While avoiding that labor may be convenient, it also makes our personal tastes flimsier, less hard-won.

在流媒体和社交媒体出现之前，文化常常让人感觉相对稀缺和有限。你要么能接触到一些东西，要么就接触不到。我对多媒体最早的记忆之一是：小时候，我和弟弟每天早上都会打开电视。工作日早上，我们想去看一部超级马里奥动画片。我们确保在录像机里装上一盘VHS录像带，因为我们想亲眼见证并捕捉某个特定的剧集，就像把收音机里的歌曲录到磁带上一样。我记得有一集特别节目讲的是马里奥从火热的地牢里救出耀西。第一次看完之后，它在我们心中就成了神话，但我们只能偶然通过有线电视节目看到它。既没办法用谷歌搜索，也没办法在线观看这部经典动画片。我们得有双重运气：一集必须在当天播出，而且必须成功录制到那盘等待的录像带上。（录像带并不总是可靠的。）只有这样，我们才能完全拥有它，随心所欲地观看超级马里奥。

Prior to streaming and social media, culture often felt relatively scarce and finite. Either you had access to something or you didn’t. One of my earliest memories of multimedia is this: As kids, my younger brother and I would turn on the television early every weekday morning to try to catch a Super Mario cartoon. We’d make sure to have a VHS tape loaded in the VCR, because we wanted to witness and capture a specific episode, like taping a song from the radio onto a cassette. One special episode, I recall, had something to do with Mario rescuing Yoshi from a fiery dungeon. After we saw it the first time, it acquired mythical status in our minds, but we could access it only by chance through cable programming. There was no way to Google it or stream the vintage show online. Our luck had to be twofold: the episode had to be on that day, and it had to record successfully onto the waiting tape. (The tape was not always reliable.) Only then would we fully possess it and be able to watch Super Mario at will.

这种体验节奏缓慢，充满摩擦。然而，正是这种坚持不懈的挑战，让我对这一集产生了如此强烈的共鸣。我为它投入了如此多的时间和情感，以至于二十多年后我仍然记忆犹新，尽管它现在被保存在一盘VHS录像带里，遗失在那栋房子的深处。对于任何通过信息流传递的普通数字内容，我都无法表达同样的感受。我们或许会在信息流中发现新事物，但我们必须抓住它们，深入挖掘，以免它们再次消失在虚无之中。我们必须有意识地与过滤世界的速度和无摩擦性作斗争。

The experience was slow and full of friction. Yet the challenge of holding on to the episode was why it acquired such a compelling aura for me. I invested so much time and emotional energy into it that I still remember it some twenty years later, although now it remains on a VHS tape lost in the bowels of that house. I can’t say the same for any piece of generic digital content delivered through a feed. We might discover things in our feeds, but we must grab on and dig deeper into them for ourselves before they disappear once more into the ether. We must consciously fight against the speed and the frictionlessness of Filterworld.

深入挖掘曾经是寻找文化的默认任务，尤其是在早期互联网上。我第一次接触动漫是九十年代末美国有线电视台上的《龙珠Z》，但当我想寻找比它无休止的青少年争吵更复杂的故事情节时，我不得不上网。我找到了一些论坛，在那里，经验丰富的动漫迷们讨论他们最喜欢的动漫，他们的目的并非像今天的平台那样为了吸引粉丝或将专业知识货币化，而是出于个人的热情。这些论坛就是“消费社群”，学术界用这个术语来描述围绕特定共同追求而聚集在网上的不同群体，无论是交换产品技巧还是讨论先锋文学。一篇论文将消费社群描述为一种“相互学习”的形式——我们共同弄清楚我们在寻找什么以及如何找到它。像推特和Facebook 的界面不稳定，算法易被操纵，不利于相互学习。

Digging deeper was once the default task of finding culture, particularly on the early Internet. My first introduction to anime was Dragonball-Z on American cable in the late nineties, but when I wanted to find more complex storytelling than its endless adolescent battles, I had to go online. I found forums where much more experienced anime fans debated their favorites, not in order to gain followers or monetize their expertise as on today’s platforms, but out of personal passion. Those forums were “communities of consumption,” a term that academics have used to describe the diverse groups of people that congregate online around a particular shared pursuit, whether swapping product tips or discussing avant-garde literature. One paper described communities of consumption as a form of “mutual learning”—we collectively figure out what it is that we’re looking for and how to find it. The likes of Twitter and Facebook, with their unstable interfaces and manipulative algorithms, are less conducive to mutual learning.

文化陷阱有很多种，算法推荐几乎可以瞬间完成。有了TikTok的推送，你从完全不知道ASMR这类刺激视频的存在，到沉浸其中，简直是捷径。一天早上，我还在床上，TikTok就开始给我推送一大堆来自丹麦美国侨民的视频日记，展示丹麦的育儿假政策和舒适的咖啡馆，我感受到了这种加速感。这些人的存在合情合理，但我不知道他们正在制作一部通俗易懂的、长达一分钟的生活纪录片。我关注了一些账号，但它们一涌而上，就又销声匿迹了，我再也找不到回到那个特定类型的频道了。没有明确的标签将它们关联起来；算法推送根据内容的参与模式将它们整合在一起，并与我自己的模式相匹配。Spotify上特定类型的音乐或Twitter上的一场争论也是如此。在那一刻，内容感觉包罗万象，但一旦你离开，它就变得毫无意义了。

There are various ways to go down a cultural rabbit hole. Algorithmic recommendations do it almost instantly. You shortcut from not knowing that ASMR sponge-squeezing videos exist to being immersed in dozens of them thanks to TikTok’s feed. I felt that sense of acceleration one morning when, while I was still in bed, the app began delivering me a slew of video diaries from American expats in Denmark, showing off the country’s parental leave policies and cozy coffee shops. It stood to reason that these people would exist, but I had no idea they were making digestible minute-long documentaries of their lives. I followed some of the accounts, but as soon as they flooded in, they also subsided, and I couldn’t find my way back to that specific genre. There was no hashtag explicitly linking them; the algorithmic feed had curated them together based on the content’s pattern of engagement, matching it to my own. The same happens for a particular genre of music on Spotify or an argument happening on Twitter. In the moment, the content feels all-encompassing, and yet it’s totally insignificant once you wander away.

更缓慢、更谨慎的方法是自己去探索这些文化脉络，并规划自己的路径，收藏账户，与其他志同道合的人交流，互相交流，就像我在动漫论坛或推特早期那样，当时它还没有变得过于庞大，难以掌控。这是一种更有意识、更有目的性的消费方式——在信息流让我们能够轻易地将在线消费的选择外包出去之前，这种消费方式是强制性的。这让人想起“鉴赏家”这个词。在艺术史的语境中，这个描述可以追溯到18世纪，当时的“鉴赏家”指的是那些仅凭观察就能辨别出作品作者的业余收藏家。他们会在研究和编目过的作品中寻找艺术家标志性的表达方式。鉴赏家主要通过消费行为来积累专业知识。德国人约翰·约阿希姆·温克尔曼就是这样一位鉴赏家；尽管他并非出身贵族，他的职业生涯始于作为一名教师，他成为了古希腊和罗马艺术最重要的收藏家和学者之一。他的努力为艺术史学科奠定了基础。

The slower and more careful approach is to seek out these seams of culture yourself and chart your path, bookmarking accounts, connecting with other people interested in the same things, and comparing notes, the way I did on anime forums or the early days of Twitter, before it became too vast to maintain a grasp on. This is a more conscious and intentional form of consumption—a form that was mandatory before feeds made it so easy to outsource our choices about what to consume online. It recalls the term connoisseur. In an art history context, the descriptor dates as far back as the eighteenth century, when connoisseurship referred to amateur collectors who could tell which artist painted a work based solely on looking at it. They sought out the artist’s signature gestures in a given work, which they had studied and cataloged. Connoisseurs developed expert knowledge, largely through the act of consumption. The German Johann Joachim Winckelmann was one such connoisseur; though he was not born into nobility and began his career as a schoolteacher, he became one of the foremost collectors and scholars of ancient Greek and Roman art. His efforts helped form the basis for the discipline of art history.

在TikTok上，成为鉴赏家更加困难，因为你几乎没有机会培养专业知识或整合你所看到内容的背景。你必须努力工作，摆脱信息流的浮夸路线，逐渐完善你所追寻的东西。这种更慢、自我管理的文化欣赏方式的好处在于，它或许能让你更好地欣赏手头的内容，你或许能够引导他人走上你走过的路，向他们展示如何欣赏同样的事物。这种方式更具可持续性，也更尊重文化，把它视为重要的东西，而不是转瞬即逝、仅仅用来吸引短暂注意力的东西。创造原创作品需要付出大量的人力，无论预期结果如何。正如我的艺术评论家朋友Orit Gat曾经半开玩笑半认真地告诉我的那样，你应该花和艺术家创作一样长的时间去欣赏一幅画。浏览信息流中如此多语无伦次的内容，我们根本没有机会去消化吸收、学习和理解，更不用说将这些理解传递给他人了。这种肤浅的消费模式导致了 Filterworld 文化的整体平淡化。

On TikTok, it’s harder to become a connoisseur because you have little chance to develop expertise or assemble the context of what you’re looking at. You must work at it, get off the slick routes of the feed, and gradually refine the thing that you seek. The benefit of the slower, self-managed approach to culture is that it might lead to a greater appreciation of the content at hand, and you might be able to lead another person down the same path that you followed, showing them how to appreciate the same things. It’s more sustainable and more respectful of culture, treating it as something important rather than ephemeral, merely fodder for brief attention spans. A lot of human effort is required to create something original, no matter the intended outcome. As my art-critic friend Orit Gat once told me, partly joking but also serious, you should look at a painting for as long as it took the artist to paint it. Flipping through so much incoherently assembled content in our feeds, we don’t have the opportunity to assimilate it, to learn and understand, much less pass on that understanding to others. That encouraged shallowness of consumption contributes to the overall flatness of culture in Filterworld.

十几岁的时候，我努力成为一名动漫鉴赏家。从老套的《龙珠》开始，多亏了论坛推荐，我逐渐接触到像《恋爱房东》和《人形电脑天使心》这样至今仍令人尴尬的浪漫喜剧动漫。这些动漫过度沉迷于男性视角（这类动漫的另一个术语是“后宫动漫”），但当时的我却觉得它们很新奇，因为它们充满了人际关系和科幻元素。我身边没有人可以谈论这些发现，只有新兴的网络社区。直到我发现了安部芳年的作品系列，比如《玲音》，我才发现自己的鉴赏能力还有很长的路要走。继《玲音》内敛的赛博朋克叙事之后，安部芳年创作了一部名为《灰羽联盟》的动漫，翻译过来就是“炭羽联盟”。这更奇怪。故事中，一个名叫拉卡的女孩从一个昆虫茧中重生，化身为某种炼狱天使，背上长出了退化的翅膀。她以这种形态出现，仿佛从一场噩梦中醒来，变成了一个田园诗般古老的这座城墙环绕的城镇，四周环绕着田野。拉卡被迫加入了一种邪教。这些被称为灰羽的天使不能拥有任何新东西，不被允许使用金钱，必须在社区中工作，并且被禁止离开城墙。

As a young teenager, I tried to become a connoisseur of anime. From the clichés of Dragonball I progressed, thanks to forum recommendations, to still-cringey romantic-comedy anime titles like Love Hina and Chobits. These were overindulgent of the male gaze (another term for the genre was “harem anime”) but felt novel to me at the time, with their interpersonal dramas and sci-fi tropes. I had no one around me to talk to about these discoveries, only the nascent online community. I found out how much further there was to go in my own connoisseurship when I discovered Yoshitoshi ABe’s series, like Lain. After Lain’s introverted cyberpunk narrative, ABe created an anime called Haibane Renmei, which translates to “Charcoal Feather Federation.” It was even stranger. In it, a girl named Rakka is reborn from an insectoid cocoon as a kind of purgatorial angel, complete with vestigial wings growing out of her back. She emerges in that form, as if from a bad dream, into a bucolic, ancient-seeming walled town surrounded by fields. Rakka has involuntarily joined a kind of cult. The angels, called Haibane, cannot own anything new, are not allowed to use money, must hold down jobs in the community, and are forbidden from leaving the town walls.

尽管剧情惨淡，这部动画却令人感到慰藉。如果它最初看起来像一场噩梦，它最终会演变成一个神话寓言。十四岁时我第一次看《灰羽联盟》时，还不知道“挽歌”这个词，但现在我会用它——悲伤、萦绕心头，并蕴含着一种令人愉悦的忧郁。动画中粗糙的线条和沉闷自然的色调增强了这种氛围。多年后，我发现这部动画的灵感来源于村上春树1985年出版的小说《世界尽头与冷酷仙境》。一种文化元素引出了另一种，我顺着这条线索继续追寻。小说和动画有着共同的氛围、词汇和哲学——它们都以城墙环绕的城镇、某种低保真蒸汽朋克美学以及通过穿越另一个世界（或许是真实世界的一部分，或许不是）来重新理解人生的叙事为特色。《灰羽联盟》是我最早发现的、真正属于我自己的文化之一——一种我默默邂逅的全新事物。我直觉地认为，我周围的大多数人未必会喜欢或理解这部作品，但它却深深地触动了我。它非常私人。

Despite the bleak scenario, the show is comforting. If it at first appears as a nightmare, it grows into a mythological fable. I didn’t know the word elegiac when I first watched Haibane Renmei as a fourteen-year-old, but it’s the one I would use now—sad and gently haunting, invested with a pleasurable melancholy. The atmosphere was enhanced by the scratchy line work of its animation and the dull, organic color palette. Years later I discovered that the show was inspired by Haruki Murakami’s novel Hard-Boiled Wonderland and the End of the World, originally published in 1985. One piece of culture led to the other, and I followed that thread. The novel and the anime share a mood, a vocabulary, and a philosophy—they both feature walled towns, a certain lo-fi steampunk aesthetic, and the narrative of comprehending your life in a new way by traveling through a different world, which may or may not be part of the real one. Haibane Renmei was one of my first cultural discoveries that truly felt like my own—something new that I encountered in silence. I intuited that most of the people around me wouldn’t necessarily like or understand the show, but it nevertheless spoke deeply to me. It was personal.

如果没有互联网，我当时根本无法接触到这样的艺术作品——不仅因为我是在互联网上发现它的，更因为互联网是唯一能找到它的地方。 2002年， 《灰羽联盟》在日本首播时，并没有立即被翻译成英文。动漫爱好者们聚集在网上，制作“粉丝字幕”——将日语DIY翻译成字幕，然后制作成动画数字文件的字幕。粉丝字幕是热情的结晶。虽然远非完美（许多字幕拼写错误，文本内容也存在问题），但它们仍然让我们这些互联网用户第一次看到了一些除非我们自己学习日语，否则无法观看的动画。这也是在线消费社区的另一个好处。

I would not have had access to such art at the time without the Internet—not only because that’s where I discovered it, but because that was the only place it was to be found. When Haibane Renmei debuted in Japan in 2002, it was not immediately translated into English. Anime aficionados congregated online and produced “fansubs”—DIY translations from Japanese that were turned into captions for digital files of the show. Fansubs were products of passion. Though far from perfect (many had terrible spelling and textual glitches), they nevertheless offered first glimpses of shows we Internet denizens couldn’t have consumed otherwise, short of learning Japanese ourselves. It was another benefit of the online community of consumption.

反过来，如果没有早期互联网的这种资源和祸害，粉丝字幕也就不会被下载。文化、文件盗版。当时没有流媒体服务，但有像 Kazaa 和 BitTorrent 这样的工具，允许用户集体上传和下载他们想要的任何媒体类型——MP3、电影、PDF。如果至少有一个人用他们的互联网连接上传某个特定文件，你就可以搜索并下载它，质量和速度各不相同。上传单个文件的用户越多，下载速度就越快，从而形成了一种松散的质量排名，没有 feed 或算法推荐。BitTorrent 称上传用户为“种子用户”，越多，文件的质量就可能越好。有时，这种质量是字面意义上的，比如分辨率或字幕。但它也是抽象的，是衡量艺术成就或文化意义的标准：更多人觉得这个特定的东西值得为分享付出成本和风险。

The fansubs, in turn, wouldn’t have been available to download without that source and scourge of so much of early Internet culture, file piracy. There were no streaming services, but there were tools like Kazaa and BitTorrent, which allowed users to collectively upload and download any kind of media they wanted—MP3s, movies, PDFs. If at least one other person devoted their Internet connection to uploading a particular file, you could search for and download it, with varying degrees of quality and speed. The more users uploading a single file, the faster it could be downloaded, leading to a loose form of quality ranking, without feeds or algorithmic recommendations. The more “seeders,” as BitTorrent called uploading users, the better the file at hand was likely to be. Sometimes that quality was literal, in terms of resolution or subtitles. But it was also abstract, a measure of artistic achievement or cultural significance: more people felt that this particular thing was worth the cost and risk of sharing it.

我记得自己曾经为了下载专辑或动漫季等上好几天甚至好几周，绿色的进度条缓缓流淌。那时电脑还未进入“永远在线”的时代；我必须确保电脑桌面保持唤醒状态，才能保证BT下载的流畅。有时，在经历了这么多工作之后，文件会被贴错标签，无论是出于错误还是恶意，我都不得不重新开始。但这个共享网络让我感到与其他粉丝的联系。这是我余生中一直缺失的东西，能够和其他人一起深入探索这种痴迷的兴趣，不受评判——事实上，是带着认同感。（当然，这其中也包含着一种刻意的反叛，一种寻求差异的尝试，而不是仅仅把自己融入到高中田径队之类的队伍中，尽管即使我尝试过，也做不到。）说我们这一代人是在互联网上长大的，这几乎成了老生常谈，但我可以说，我是在互联网上第一次意识到自己是一个成年人，一个独立于我周围环境的存在。在那里，我可以从各种不同的灵感中构建自我。如此亲密地接触文化，几乎将其融入自我，或许是我们青少年时期最擅长的事情。

I remember waiting days and even weeks for albums or anime seasons to fully download, the green progress bar slowly filling. This was before the era of always-on computers; I had to make sure the desktop stayed awake to keep BitTorrent chugging along. Sometimes, after all that work, the files would be mislabeled, whether by mistake or malice, and I would have to start over again. But the sharing network made me feel connected to these other fans. It was something I was missing in the rest of my life, the chance to delve into this obsessive interest with other people and without judgment—in fact, with approval. (Of course, there was also an element of intentional rebellion, a seeking of difference by not just casting myself in with, say, the high school track team, though I couldn’t have done that if I tried.) It’s almost a cliché to note that my generation grew up on the Internet, but I can say that I first became aware of myself as an adult person on the Internet, a being separate from my immediate context. It was a space in which I could construct myself out of the disparate inspirations I found there. Engaging so intimately with culture, almost ingesting it into the self, is an activity that we are perhaps most capable of as teenagers.

刘易斯·海德（Lewis Hyde）在其1983年出版的《礼物》（The Gift）一书中，将艺术品定义为艺术家通过创作行为自由赠予的物品，无论其最终归于何处：“一件艺术品蕴含着艺术家礼物的精神。” 但从某种程度上来说，品味也可以是一种礼物。引入它无需任何成本。将你认为某人可能会喜欢的文化元素传递给他们，并且这一行为可能使所有相关方受益。毕竟，文化并非一对多的广播系统，而是一个像BitTorrent一样的点对点网络，我们通过有意识地分享来共同决定什么对我们最有意义。正如海德所写：“礼物的精神因持续的捐赠而永存。”

In his 1983 book The Gift, Lewis Hyde defines artwork as something freely given by the artist through her creative act, no matter where it ends up: “a work of art contains the spirit of the artist’s gift.” But in a way, taste can be a gift, too. It costs nothing to introduce someone to a new piece of culture that you think they might like, and the act might benefit all parties involved. Culture, after all, is not a one-to-many broadcast system but a peer-to-peer network, like BitTorrent, where we collectively determine what means the most to us by intentionally sharing it. As Hyde wrote, “The spirit of a gift is kept alive by its constant donation.”

我的音乐之路也曾有过类似的“兔子洞”式经历。我从母亲那里继承了对戴夫·马修斯乐队的欣赏，那是九十年代混乱的原声即兴乐队的巅峰之作，我无法完全原谅，但也无法抹去。我在离家一小时车程的庞大的Borders书店买CD，也正是在那里，我找到了一批新近出版的日文翻译漫画。但在网上，我却走了一条更迂回的路来寻找我的音乐品味。

I had a similar rabbit-hole experience with music. I inherited from my mother an appreciation for Dave Matthews Band, the pinnacle of shambolic nineties acoustic jam bands, that I can’t totally excuse but also can’t erase. I bought CDs at the monolithic Borders outlet an hour’s drive from my house, the same place I found a newly available crop of translated Japanese manga. But online I traveled a more circuitous route to find my taste in music.

我第一个真正的归宿是一个大型多人在线角色扮演游戏（MMORPG）的论坛——一个你可以通过互联网与成千上万的玩家同时共处的游戏世界——里面有一个辅助音乐版块。后来我转到了一个专门讨论戴夫·马修斯乐队的论坛，叫做 AntsMarching.org（再次让我感到尴尬）。最后，在21世纪初，我加入了一个稍微宽泛一些的独立音乐论坛，叫做 UFCK.org，那里聚集了一些对乐队有些不满——或许是已经不再喜欢乐队了——的 DMB 乐迷。

My first real home was a forum for a massive multiplayer online role-playing game (MMORPG)—a video-game world that you inhabit with thousands of other players at the same time over the Internet—that had an ancillary music section. Then I moved to a devoted Dave Matthews Band forum called AntsMarching.org (again, cringe). Finally, over the course of the early 2000s, I landed at a slightly broader indie music forum called UFCK.org, which gathered DMB fans who had become somewhat disaffected with the band—grown out of it, perhaps.

就像考古学家清除几个世纪的尘土一样，如今我在网上只能找到UFCK的些许踪迹，大多是2006年它关闭后在其他论坛上发布的回忆或“难民”主题帖。但在当时，它是我网络生活的中心，也是我找到其他乐队和音乐家推荐的地方，这些乐队和音乐家后来成了我品味的核心：安德鲁·伯德（Andrew Bird）、十二月党人（The Decemberists）、苏菲扬·史蒂文斯（Sufjan Stevens）。这种选择远非独一无二；它成为了北美那些过于沉迷网络或过于嬉皮的青少年的一种普遍的音乐品味，是跨越地域的网络同质化的早期象征。但在我现实生活中，这却很不寻常。我认识的其他人中没有谁听过这些乐队的音乐。这些论坛也充当着音乐会录音的档案馆，尤其是DMB的。录音师会带着录音设备——麦克风、吊杆和录音机——来到音乐会现场，然后在网上分享录音成果。论坛成员们会讨论哪场DMB音乐会或哪场单场现场演出。歌曲表演是最好的，这可以追溯到 1991 年乐队成立之初。集体讨论在音乐欣赏的这个小圈子里形成了一种 DIY 经典——遵循 Grateful Dead 和 Phish 早期建立的生态系统。

Like an archaeologist brushing away centuries of dirt, I can find only bare traces of UFCK online now, mostly remembrances or “refugee” threads on other forums after it shut down in 2006. But at the time, it was the center of my online life, and it was where I found the recommendations of other bands and musicians that became central to my taste: Andrew Bird, the Decemberists, Sufjan Stevens. This selection was far from unique; it became a kind of generic musical taste for North American teenagers who were too online or too hipster, an early symbol of online homogeneity cutting across geography. But in my real-life surroundings, it was unusual. I didn’t know anyone else who was listening to these bands. The forums also functioned as archives for concert recordings, particularly for DMB. Tapers would bring recording equipment—microphones, booms, and tape decks—to concerts, then share the result online. Members of the forums debated which DMB concert or single live performance of a song was best, stretching back through to the band’s formation in 1991. The collective discussion formed a DIY canon in this small subset of music appreciation—following an ecosystem established earlier by the Grateful Dead and Phish.

随着时间的推移，我通过文件共享收集了自己的音乐会MP3数字档案，并按年份排序，就像品酒师收藏陈年酒窖里的陈年佳酿一样。我开始最欣赏九十年代中期的录音，并认为1997年是这支乐队最好的一年。录音的静态感和残缺，以及演出的另类，都无关紧要。有意识地积累收藏，并思考自己最欣赏某个创作者或文化群体的哪些方面，意味着成为一名鉴赏家。“鉴赏家”这个词听起来有些自命不凡，但我们可以成为任何事物的鉴赏家：真人秀、噪音音乐、苹果派食谱。我们通过算法推送获得的可用性——可以即时访问各种素材并随意浏览——却失去了鉴赏能力，这需要深度和用心。归根结底，这是一种深刻的欣赏，既要欣赏艺术家的作品，也要欣赏我们自身的品味。

Over time, I gathered my own digital archive of concert MP3s via file sharing, sorted by year, the way an oenophile keeps a wine cellar of aged bottles. I started to appreciate mid-nineties recordings the most, staking out an opinion that 1997 was the band’s best year. It didn’t matter that the recordings were staticky and incomplete, or the performances unorthodox. Having to consciously accrue a collection and think through what you enjoy most about a particular creator or body of culture means becoming a connoisseur. The term has a pretentious quality, but we can become connoisseurs of anything: reality TV shows, noise music, apple pie recipes. What we gain with algorithmic feeds in terms of availability—having instant access to a broad range of material to be scanned at will—we lose in connoisseurship, which requires depth and intention. It’s ultimately a form of deep appreciation, for what the artist has done as well as the capacities of our own tastes.

从经济角度来看，我青少年时期的那种盗版行为是不道德的，更不用说大多数情况下是非法的。我知道我偷的是那些在美国获得商业授权的专辑和电视节目。艺术家们并没有从我的粉丝群体中获得经济利益。（这个问题并不新鲜；19世纪的小说家面临着外国出版商的盗版，就像2000年代的音乐家面临着Napster的盗版一样，历史学家奥兰多·菲吉斯在其著作《欧洲人》中对此有过令人难忘的描述。）但作为2000年代文化传播的典范，我认为没有什么比论坛和文件共享的生态系统更好了。它就是我的亚历山大图书馆。互联网早期是一个媒介较少的口口相传的环境，专业知识从一个人传到另一个人，慢慢地建立起群体和社区，最终凝聚成一个整体的、但仍然可以理解的文化。

Economically, the kind of piracy I was practicing as a teenager was unethical, not to mention most often illegal. I knew that I was stealing these albums and television shows, the ones that had been commercially licensed in the United States. The artists weren’t benefiting financially from my fandom. (This problem was nothing new; novelists in the nineteenth century faced pirated editions of their books from foreign publishers just as musicians in the 2000s confronted piracy from Napster, as the historian Orlando Figes memorably described in his book The Europeans.) But as a model for cultural distribution, as it existed in the 2000s, I can think of nothing better than the ecosystem of forums and file sharing. It was my Library of Alexandria. The earlier era of the Internet was a much less mediated environment of word of mouth, with specialized knowledge filtering from one person to another, slowly building up to groups and communities, cohering into a holistic culture that was nevertheless graspable.

我必须承认，我对这段时光的敬仰部分源于怀旧。我们都梦想着那段自我相对来说还未成型，与艺术的邂逅有着惊人的力量。青少年更容易接受新体验，无论他们使用什么技术来获取这些体验，他们也更倾向于沉迷于艺术，成为鉴赏家。但我意识到，我非常欣赏这些在线互动，因为它们建立在人与人之间的推荐之上，而不是自动化的推荐。必须有人足够关心地告诉我他们喜欢什么，而我也必须足够关心地信任他们并公平地尝试。这种文化推荐——传达认可——是社会和道德行为。我们告诉彼此我们喜欢​​的东西，就像蜜蜂通过舞蹈来传递一朵特别繁茂的花朵的位置一样。正是这种行为将我们联系在一起。

I must admit that my admiration for this period is driven in part by nostalgia. We all dream of that period of youth when the self is relatively unformed and encounters with art have a staggering power. Teenagers are more open to new experiences, regardless of what technology they use to consume them, and have the tendency and time to indulge obsessions, to become connoisseurs. But I’ve realized that what I appreciated so much about those online interactions is that they were built on person-to-person recommendations, not automated ones. Someone had to care enough to tell me what they liked, and I had to care enough to trust them and give it a fair try. Such cultural recommendations—communicating approval—are social and moral acts. We tell each other that we like things the same way that bees perform dances to pass on the location of a particularly fruitful flower. The very act brings us together.

那么，推荐系统是这种交换的更抽象的版本。在其中，我们的网络行为由算法聚合，然后进行处理和平均，并输出到外部，以创建强加于他人的消费模板。它表面上加速了这一过程，实际上却阻碍了文化的有机发展，反而优先考虑扁平化和千篇一律，而这些审美在数字平台的网络上最容易传播。从某种程度上来说，这本书试图重新定义推荐系统。我们应该更多地谈论我们喜欢的事物，共同体验它们，并建立我们自己精心打造的喜好和厌恶的集合。这并非为了微调算法，而是为了我们集体的满足。

Recommender systems, then, are a much more abstracted version of this exchange. Within them, our net behaviors are aggregated by an algorithm, then crunched and averaged and spit back out to create templates of consumption that are imposed on other people. In the guise of speeding it up, it actually impedes that organic development of culture and instead prioritizes flatness and sameness, the aesthetics that are the most transmissible across the networks of digital platforms. In a way, this book is an attempt to recapture recommendations from recommender systems. We should talk even more about the things we like, experience them together, and build up our own careful collections of likes and dislikes. Not for the sake of fine-tuning an algorithm, but for our collective satisfaction.

毕竟，推荐是人类的一项专业工作。有些人致力于探索我们应该接触哪些文化，以及我们可能会欣赏哪些文化，并根据时局调整方法，拓展品味的界限。你可能会在精品店、艺术博物馆、广播电台，甚至电影院的幕后发现他们。这些专业的推荐人被称为“策展人”。他们确保推荐的内容物有所值。他们进行语境化，向我们介绍新鲜事物，并给予我们足够的挑战，避免我们趋同。他们引导我们的消费。尽管“策展人”这个词在互联网上可能被过度使用，但我们真正需要的是更多的“策展”——培养和运用个人品味。

Recommending things is a professional human job, after all. There are people who work to figure out what culture we should be exposed to and what we might appreciate, adapting their approaches to the moment and expanding the boundaries of what is considered tasteful. You might find them in a boutique, at an art museum, on a radio station, or behind the scenes at a movie theater. These professional recommenders are called curators. They make sure that exposure goes to what merits it. They perform the labor of contextualization and introduce us to what’s new, challenging us enough that we avoid homogeneity. They guide our consumption. Though the word might get overused on the Internet, what we really need is more curation—the cultivation and deployment of personal taste.

策展的力量

THE POWER OF CURATION

管理始于责任。根据 1875 年的一本词典， curatore一词的词源原意是指古罗马的“公职人员”，这些职位早于公元前 27 年开始统治的奥古斯都皇帝。他们管理城市维护的各个方面：有台伯河的管理员，负责食品采购、将水输送到城市的渡槽以及举办公共运动会。在拉丁语中，curare的意思是照顾，而curatio则表示关注和管理。几个世纪以来，这个词的含义变得不再世俗，而更加精神化，但仍然与照料有关。到了 14 世纪，名词curate指的是担任宗教指导的人。在 1662 年的英语版《公祷书》中，助理牧师是教区的副牧师，负责指导他的教区居民并照顾他们灵魂的“治疗”——照料。至少从十九世纪中叶开始，“策展人”就专门描述了博物馆及其藏品（无论是艺术品还是历史文物）的管理者——物品的管理者，而不是人的管理者。

Curation begins with responsibility. The etymological ancestor of the word curatore was a term for ancient Roman “public officers,” according to an 1875 dictionary, positions that predated the emperor Augustus, whose reign began in 27 BCE. They managed various aspects of the city’s upkeep: there were curatores of the Tiber River, the purchase of foodstuffs, the aqueducts that carried water into the city, and the hosting of public games. In Latin, curare meant to take care of, and curatio indicated attention and management. Over the centuries, the word’s meaning became less mundane and more spiritual, but still related to caretaking. By the fourteenth century, the noun curate referred to a person working as a religious guide. In the 1662 English-language Book of Common Prayer, a curate was the deputy priest of a parish, guiding his parishioners and looking after the “cure”—care—of their souls. Since at least the mid-nineteenth century, curator has specifically described the manager of a museum and its collections, whether artworks or historical artifacts—a steward of objects instead of people.

策展一词的词源暗示了其重要性，它不仅仅是一种消费行为、品味展示，甚至自我定义，更是对文化的呵护，一个严谨而持续的过程。二十世纪下半叶，名人策展人崛起，他们是极具影响力的潮流引领者，他们的选择影响着各自时代的集体品味。从1932年开始，现代艺术博物馆首任建筑策展人菲利普·约翰逊（其家族财富也资助了该部门）是推广现代主义设计美学的重要人物之一。约翰逊将路德维希·密斯·凡德罗等人设计的朴素工业家具在博物馆空间中展示，这些家具起初令人震惊，但约翰逊逐渐使其变得易于接受。20世纪60年代，比利时出生的亨利·盖尔扎勒在大都会博物馆担任美国艺术策展人，逐渐将目光转向在世艺术家，这在当时对于这样的机构来说实属罕见。盖尔扎勒利用自己的职位支持早期波普艺术时代的艺术家，包括他的密友安迪·沃霍尔和大卫·霍克尼，以及罗伯特·劳森伯格和贾斯珀·琼斯等艺术家。（策展人作品波普艺术总是有些个人化，但盖尔扎勒比大多数人走得更远。）波普艺术的艳丽和平庸的特质最初让观众感到震惊，但策展人帮助将其置于特定背景中，论证其重要性，直到 1977 年盖尔扎勒离开大都会艺术博物馆时，这场运动才被广泛接受为艺术史经典的一部分。

The word’s etymology hints at the importance of curating, not just as an act of consumption, taste displaying, or even self-definition, but as the caretaking of culture, a rigorous and ongoing process. The second half of the twentieth century saw the rise of celebrity curators, powerful tastemakers whose choices influenced the collective tastes of their eras. Beginning in 1932, Philip Johnson, the Museum of Modern Art’s first curator of architecture (a department that his family wealth also funded), was one of the figures responsible for popularizing the aesthetic of modernist design. By showing off the stark industrial furniture of Ludwig Mies van der Rohe and others, which appeared shocking at first, in the space of a museum, Johnson gradually made it palatable. In the 1960s, the Belgian-born Henry Geldzahler worked as a curator of American art at the Metropolitan Museum, gradually focusing on living artists, which was rare at that time for such an institution. Geldzahler used his position to support the early Pop era of artists, including Andy Warhol and David Hockney (his close friends), as well as artists like Robert Rauschenberg and Jasper Johns. (A curator’s work is always somewhat personal, but Geldzahler took it further than most.) The garish and banal qualities of Pop art initially shocked viewers, but the curator helped to contextualize it, arguing for its importance until the movement became widely accepted as part of the art-historical canon by the time Geldzahler left the Met in 1977.

正如博物馆研究学者布鲁斯·阿特舒勒（Bruce Altshuler）在其1994年出版的《展览中的先锋派》（The Avant-Garde in Exhibition）一书中所言，那几十年见证了“策展人作为创造者的崛起” 。二十世纪末二十一世纪初，我们见证了一批明星策展人的涌现，其中包括汉斯·乌尔里希·奥布里斯特（Hans Ulrich Obrist）和卡罗琳·克里斯托夫-巴卡基耶夫（Carolyn Christov-Bakargiev）。他们如同无官职的外交官，在国际范围内运作，跨越不同的机构，最终落脚于博物馆或画廊，策划既能反映个人情感又能体现当下的展览。策展实践成为一种在国际范围内确立和恢复知名度的策略。尼日利亚策展人奥奎·恩维佐（Okwui Enwezor）策划了多项重要展览，例如1989年的“大地魔术师”（Magicians de la Terre）和2002年的“文献展”（Documenta），后者是每五年举办一次的德国当代艺术展。在这些展览中，恩维佐将来自被忽视的国家和背景的艺术家与西方艺术界已经成名的艺术家放在一起，有力地论证了当代艺术是全球思想交流的一种方式——德国艺术家安瑟姆·基弗与中国艺术家黄永砯和澳大利亚土著艺术家约翰·马沃德朱尔并列。

Those decades saw “the rise of the curator as creator,” as the museum-studies scholar Bruce Altshuler put it in his 1994 book The Avant-Garde in Exhibition. In the late twentieth century and the early twenty-first, we have seen a set of celebrity curators emerge, including Hans Ulrich Obrist and Carolyn Christov-Bakargiev. They operate on an international scale, like diplomats without portfolio, and work across different institutions, alighting at a museum or gallery to organize exhibitions that reflect both their personal sensibilities and the current moment. Curatorial practice became a strategy for asserting and restoring visibility on an international scale. The Nigerian curator Okwui Enwezor organized major exhibitions like the 1989 Magicians de la Terre and 2002 Documenta, a German survey of contemporary art that happens every five years. In those exhibitions, Enwezor placed artists from overlooked countries and backgrounds alongside already famous names from the Western art world, making a powerful argument for contemporary art as a global exchange of ideas—the German Anselm Kiefer next to the Chinese Huang Yong Ping and the Australian indigenous artist John Mawurndjul.

从某种意义上说，明星策展人与推荐算法截然相反：他们运用自身所有的知识、专业技能和经验，以极致的敏感性和人性化，来决定向我们展示什么以及如何呈现。鉴于他们的职业范围——这确实是世界上最酷的职业之一——策展人这一形象成为一种光鲜亮丽的典型，并成为其他人争相效仿的标签，也就不足为奇了。甚至连机器也在努力成为策展人。

In a sense, the individual star curators are the opposite of recommendation algorithms: they utilize all of their knowledge, expertise, and experience in order to determine what to show us and how to do it, with utmost sensitivity and humanity. Given the scope of their careers—truly one of the world’s coolest jobs—it is no wonder that the figure of the curator has become a glamorous archetype and a label that others want to latch on to. Even machines are trying to become curators.

从2010年代初开始，持续了十年，至今仍未停止，“策展”一词无处不在。它曾经是一个只有专家和学者才会使用的神秘术语，如今却被广泛地应用于Instagram帖子、化妆品配色方案以及时尚品牌配饰。以下是这只是我观察到的过度使用策展的几种情况：有影响力的人“策划”她为其创建赞助内容的公司，或策划活动的来宾名单。餐馆老板策划酒吧菜单，或食品市场策划占据其摊位的供应商。酒店精心挑选房间，仿佛每间客房都是自己独特的体验。流媒体服务策划其提供的内容。据 NPR 报道，音乐家甚至“策划了一份非传统的职业”。在社交网络时代，我们都必须策划我们的身份，即选择在个人资料页面上最能代表我们的内容。在 2012 年的一次演讲中，网络艺术家乔纳森哈里斯总结了这种转变：“策展正在取代创作成为一种自我表达的方式。”与其说是一种全身心的照顾，不如说是一种自恋的行为。

Beginning in the early 2010s and extending over the course of the decade, with no sign of stopping today, the word curated became ubiquitous. Once an arcane term used only by specialists and academics, it is now applied wantonly to grids of Instagram posts, cosmetics color palettes, and fashion brand accessories. Here are only a few of the excessive usages of curation that I’ve observed: An influencer “curates” the companies she creates sponsored content for, or curates the guest list of an event. A restaurateur curates a bar menu, or a food market curates the vendors occupying its stalls. A hotel curates its selection of rooms, as if each one were its own unique experience. A streaming service curates the content it offers. A musician even “curates an unconventional career,” according to NPR. In the social network era, we’ve all had to curate our identities, in the sense of selecting which pieces of content best represent us on a profile page. In a 2012 lecture, the Internet artist Jonathan Harris summarized the shift: “Curation is replacing creation as a mode of self-expression.” Rather than a form of committed caretaking, it more appears as an act of narcissism.

这个词的氛围充满渴望，意在赋予它意义和品质，就像干邑白兰地酒瓶上“非常特别”的标识一样。难道你不想拥有精心策划的东西，而不是未经策划的东西吗？如今，这个词仅仅表示在一组选项中进行选择，而这个决定是由一个拥有一定专业知识和意图的人做出的。然而，算法推荐也常常被描述为“策划”信息流，即使其背后没有任何意识。在我写这篇文章的时候，我的手机会自动“策划”我拍摄的照片库，将它们整理成理论上有意义且引人入胜的亮点。（至于具体采用什么标准，它没有说明。）这个词的语义饱和是因为算法的盛行使得在事物之间做出选择变成了一种新鲜事。当我们不必做出选择时，做出选择，或者知道别人做了选择，就成了一种奢侈，尽管这种奢侈有些可悲。如果我们从当代的定义推断，在信息流时代之前，一切都必须是经过策划的，从有线电视上的电视节目到广播中的歌曲。

The aura of the word is aspirational, meant to impart a sense of significance and quality, like a “Very Special” designation on a bottle of cognac. Wouldn’t you want something that is curated over something that is not? Lately, the word simply indicates a choice between a group of options, a decision that was made by a human with some semblance of expertise and intention. Yet algorithmic recommendations are also often described as “curating” a feed, even though there is no consciousness behind them. As I write this, my phone offers an automatically “curated” library of the photos I’ve taken, sorting them into theoretically meaningful and attractive highlights. (By what metric, it doesn’t say.) The semantic satiation of the word has happened because the prevalence of algorithms has turned choices between things into a novelty. When we don’t have to make a selection, doing so, or knowing that someone else did, becomes a kind of luxury, albeit a pathetic one. If we extrapolate from that contemporary definition, before the era of feeds, everything must have been curated, from the television shows on cable to the songs on the radio.

过度使用“策展人”一词，反而忽略了策展人本身的形象。策展人的责任在于做出明智的选择，呵护她所负责的藏品。如果你在高中时问我，我的人生目标是什么，我会说，我想成为一名策展人。我的愿望源于现代艺术史书籍和参观曼哈顿现代艺术博物馆的经历，我会乘坐通勤列车往返于……我每年都会去康涅狄格州几次。我参加了奥尔德里奇当代艺术博物馆的一个课外活动，馆长们会和学生们讨论他们正在合作组织展览和完成委托的艺术家。

What is lost in the overuse of the word is the figure of the curator herself, a person whose responsibility it is to make informed choices, to care for the material under her purview. If you had asked me in high school what I wanted to do with my life, I would have said that I wanted to be just such a curator. My desire was inspired by books on the history of modern art and visits to the Museum of Modern Art in Manhattan, which I would take the commuter train to from Connecticut a few times a year. I attended an after-school program at the Aldrich Contemporary Art Museum, where the staff curators would talk to students about the artists with whom they were working to organize shows and complete commissions.

高中时，我曾在某种程度上设想策展人的职责主要是决定白色画廊墙上物品的摆放位置，就像一种极端的室内装饰。我猜想他们会精确地选择每幅画作及其周围作品的摆放位置，设计展览中作品的呈现顺序。某种程度上，我是对的——摆放艺术品是这份工作很重要的一部分。但这只是它最公开的一面。大学期间，我在波士顿美术馆当代艺术部做过暑期实习，我意识到这项工作远比这更艰巨。这也是一份充满学术性、近乎苦行僧的工作，就像那些神职人员一样。这些布局决策背后是数百小时的研究、写作、思考和维护。这份工作并不一定光鲜亮丽：我大部分的实习时间都在更新实体艺术品存放位置的卡片目录。但我最自豪的时刻是在多年后，当时我意识到我为菲利普·加斯顿的一幅巨幅画作撰写的几句标签文字被陈列在博物馆里，我看到好奇的参观者走到标签前，阅读我的文字。我在某种程度上参与了那个公共策展过程，并负责向他们传达关于加斯顿和这幅画作他们需要知道的一切。

In high school, I had envisioned, on some level, that a curator’s role was principally to decide where things went on white gallery walls, like an extreme version of interior decorating. I figured they chose precisely where to place each painting and its neighboring works of art, designing the order in which the works in the show should be experienced. I was right, in a way—arranging artwork is a big part of the job. But this is only its most public aspect. When, in college, I did a summer internship at the contemporary department of Boston’s Museum of Fine Arts, I realized that the labor went much deeper. It was also a scholarly, almost ascetic job, like that of the priestly curates. Behind those placement decisions were hundreds of hours of research, writing, thinking, and maintenance. It wasn’t necessarily glamorous: I spent most of my internship time updating a card catalog of where physical artworks resided in storage. But my proudest moment came years later, when I realized that a few sentences of label text I had written for a massive Philip Guston painting had been installed in the museum, and I saw curious visitors going up to the label and reading my words. I had participated, in a small way, in that public curatorial process and was responsible for telling them what they needed to know about Guston and the painting.

互联网上或许策展泛滥，但策展却远远不够。就内容的长期管理、组织和语境化而言，所有这些流程都外包给了算法。为了探究策展人在“过滤世界”算法僵局中扮演的角色，我采访了保拉·安东内利 (Paola Antonelli)，她于 1994 年加入纽约现代艺术博物馆 (MoMA)，现任建筑与设计部高级策展人兼研发总监。安东内利是我们这个时代最具创新精神的策展人之一，我很荣幸能与她相识十余年，就艺术、设计、科技以及文化的未来展开了一场漫谈。

The Internet might have an overflow of curation, but it also doesn’t have enough of it, in the sense of long-term stewardship, organization, and contextualization of content—all processes that have been outsourced to algorithms. To determine the role of curators in the algorithmic gridlock of Filterworld, I met up with Paola Antonelli, a curator who joined the Museum of Modern Art in 1994 and is now the senior curator of its Department of Architecture and Design as well as a director of research and development. Antonelli is one of the most innovative curators of our time, and I’m lucky to have known her for more than a decade, carrying on a meandering conversation about art, design, technology, and the future of culture.

安东内利出生于意大利撒丁岛，并于米兰。在安东内利入职纽约现代艺术博物馆后，她领导了博物馆设计收藏的扩展工作，这项工作始于菲利普·约翰逊时代，并收藏了一系列乍一看似乎不属于博物馆的非传统藏品。2010年，在安东内利的领导下，博物馆获得了@符号（由于属于公共领域，因此是免费的），2012年，博物馆又购入了一批代表电子游戏历史的十四款电子游戏，其中包括《俄罗斯方块》、《 神秘岛》和《模拟人生》。

Antonelli was born in Sardinia, Italy, and studied architecture in Milan. After landing at MoMA, she led the expansion of its design collection, which began back in Philip Johnson’s day, and collected a series of unorthodox objects that may not initially seem to belong in a museum. In 2010, under Antonelli’s charge, the museum acquired the @ symbol (which was free, since it was in the public domain) and in 2012 it acquired a batch of fourteen video games that were representative of the medium’s history, including Tetris, Myst, and The Sims.

安东内利认为，世界充满了设计，从街上的消防栓到你用来打字的键盘。她的策展实践重新诠释了这些日常物品的语境，并凸显了它们的创作天赋。在她2011年的展览“与我对话”中，她收集的作品探讨了“人与物之间的交流”。“物会与我们对话，”她写道。展览涵盖了从可以用来编写代码的实体积木到为无法体验痛经的人模拟痛经的设备，以及一台1999年的纽约地铁卡自动售货机。展览的布置并非现代艺术展览通常的简朴模式，而更像是一个令人兴奋的零售商店，作品陈列在展厅中央亮橙色的货架模块上。

Antonelli makes the argument that the world is full of design, from a fire hydrant on the street to the keyboard you type with. Her curatorial practice recontextualizes these everyday objects and highlights the genius of their creation. In her exhibition “Talk to Me” from 2011, she gathered pieces that addressed “the communication between people and things.” “Things talk to us,” she wrote. The show ranged from physical blocks that could be used to write code to a device that simulated menstrual pain for those who can’t experience it and an actual New York City MetroCard vending machine from 1999. The show was installed not in the usual spartan mode of modern art exhibitions, but almost as an exciting retail store, with pieces displayed on bright orange shelf modules in the middle of the floor.

我记得当我看到它时，被其布局的能量所震撼。安东内利的三维物体拼贴画唤起了一个互动科技的世界，在这个世界中，数字和物理设备不断调节着我们的人际关系。安东内利以朴实无华却又严谨的方式排列着作品，蕴含着一整套哲学思想。事实上，算法信息流甚至可以作为展览中的展品。

I remember being shocked by the energy of the layout when I encountered it. Antonelli’s three-dimensional collage of objects evoked a world of interactive technology, in which devices both digital and physical are constantly mediating our interpersonal relationships. An entire philosophy was encoded in the unpretentious and yet rigorous way that Antonelli ordered her selections. Algorithmic feeds, in fact, could fit in as objects in that exhibition.

在约定的日期，我乘坐美铁列车从华盛顿特区前往纽约市，穿过中城区，穿过我十几岁时去纽约现代艺术博物馆时走过的那些街区，路过熟悉的公共雕塑和路边的清真餐车，我经常在那里吃午饭，然后才到达博物馆。这段旅程对我来说永远是一场朝圣之旅：我感觉自己就像一个中世纪的农民前往大教堂。正是在那里，我了解了我所热爱的艺术，也看到了我在书中读到的艺术作品，比如毕加索那幅极其抽象的《亚维农少女》，如今就静静地悬挂在画廊里。纽约现代艺术博物馆的整体馆藏涵盖约二十万件藏品，无论实体机构规模如何庞大，都无法全部展出。因此，数十位策展人不断抉择，决定展出哪些藏品，以及将它们摆放在哪里。

On the appointed day, I took the Amtrak train from D.C. to New York City and walked through Midtown, the same blocks I had traversed going to MoMA as a teenager, past familiar public sculptures and sidewalk halal carts where I had often eaten lunch before arriving at the museum. The journey is always a pilgrimage to me: I feel like a medieval peasant traveling to a cathedral. It was where I learned about so much art I love and where I could see the artworks I read about in books, like Picasso’s radically abstracted Les Demoiselles d’Avignon, casually hanging in a gallery. MoMA’s overall collection includes around two hundred thousand objects, and however large the physical institution becomes, it could never display them all. So its dozens of curators are constantly making choices about what to show and where to put it.

我在博物馆员工翼楼外遇见了安东内利。策展人通常都很时尚（擅长挑选物品），但安东内利却一改往日的黑色套装，换上了翻领带白条纹的红色衬衫和与之相配的半身裙。多年前，她就曾总结，策展人的最佳状态是“值得信赖的向导”。“他们在某些方面很专长，”她说，无论是超市里的橄榄油，还是20世纪60年代的美国绘画。更重要的是，策展是双向的：策展人必须时刻关注观众，了解他们的看法、反应，甚至情绪状态。“这可以说是一种表演艺术，”安东内利告诉我。“你赢得的是你的明星、你的信任和你的信誉，一旦你赢得了，你就必须保持它。”展览开幕后，表演并不会结束。

I met Antonelli outside of the museum’s staff wing. Curators are often fashionable (adept at choosing things), but Antonelli had exchanged the typical black ensemble for a red blouse with white stripes on the lapels and a matching skirt. Many years ago, she had concluded that curators at their best are “trusted guides.” “They are specialized in something,” she said, whether that’s olive oil in a supermarket or American paintings from the 1960s. What’s more, curation is a two-way street: one always must be aware of the viewers, their perceptions, reactions, and even emotional states. “It’s a performance art of sorts,” Antonelli told me. “You earn your stars, your trust, and your credibility, and once you’ve earned it, you have to keep it.” The performance doesn’t end once the exhibition opens.

策展人也必须尊重观众独立思考的能力。安东内利试图让她的展览主题保持开放，“90% 是成熟的”。剩下的 10% 则留给观众空间，让他们将自身的体验融入作品，自行完善想法或论证。当一切都过于预设或墨守成规时，观众就会感到疏离，因为他们缺乏自主性。“我相信我的工作不是告诉人们什么是好的，什么是坏的，而是激发他们自身的批判意识，”她继续说道。正如厨师的开胃小菜能唤醒食欲，让人们更好地欣赏接下来的菜肴一样，策展人的选择也能激发我们的感官，去思考眼前的事物。这种整体的敏感度是算法推送无法复制的。

Curators must also respect their audience’s capacity to think for themselves. Antonelli tries to leave her exhibition theses open-ended, “90 percent baked.” The remaining 10 percent gives space to the audience to bring their own experience to the work, completing the idea or argument on their own. When things are too predetermined or set in a template, audiences are alienated, because they feel no agency. “I believe that my job is not to tell people what’s good and what’s bad, but rather, it’s to stimulate their own critical sense,” she continued. Just as a chef’s amuse-bouche wakes up the appetite so one can better appreciate the meal that follows, the curator’s selection stimulates our senses to consider what’s in front of us. This kind of holistic sensitivity is something an algorithmic feed is incapable of replicating.

为了解释她的方法，安东内利带我走进了216号展厅。这个展厅融合了设计、建筑和一些视觉艺术元素，展出了许多安东内利亲自为纽约现代艺术博物馆收藏的作品。她选择了艺术家兼研究人员凯特·克劳福德（Kate Crawford）和弗拉丹·乔勒（Vladan Joler）的作品作为展厅的压轴作品。这幅名为《人工智能系统解剖图》（Anatomy of an AI System）的作品是一幅黑色背景的信息图，放大到墙面大小，记录了亚马逊Echo设备中各种形式的人力、基础设施和数据。其中，安东内利选择了“提取”这个主题。在画廊的第一个角落，三件作品碰撞在一起：在《人工智能系统解剖图》旁边安装了一个皮革模型，它类似于一头俯卧的牛，但却是一个功能性的长凳，还有一个谷歌地图上熟悉的位置图标，打印出来大约六英尺高。

To explain her approach, Antonelli walked me into gallery 216, which mingles design and architecture with some visual art, featuring many pieces that Antonelli had acquired for the MoMA collection herself. She chose a piece by the artists and researchers Kate Crawford and Vladan Joler to anchor the room. Called Anatomy of an AI System, it’s an infographic on black background expanded to the size of a wall that documents the various forms of human labor, infrastructure, and data that go into Amazon’s Echo device. From it, Antonelli took up the theme of “extraction.” In the first corner of the gallery, a trio of pieces collide: alongside Anatomy of an AI System are installed a leather form that resembles a prone cow but is a functional bench, and a location icon instantly familiar from Google Maps, printed some six feet tall.

这头牛是茱莉亚·罗曼 (Julia Lohmann) 的作品“瓦尔特劳德”牛凳。它的无头造型既令人毛骨悚然又可爱，制作它的闪闪发光的棕色皮革取自一头曾经活着的牛，然后再次塑造成牛的形状。“这头牛的存在是为了提醒人们。它几乎就像一个接地元素，”安东内利解释道。它代表着对资源（动物和环境）的提取，以创造出稀有的物体。然而，它仍然是一个物体；孩子们有时会爬上去，警卫必须冲进来。谷歌地图图钉图标由延斯·艾尔斯特鲁普·拉斯穆森 (Jens Eilstrup Rasmussen) 于 2005 年设计，应安东内利的要求，谷歌将其赠送给了纽约现代艺术博物馆。她决定将其打印得比屏幕上显示的大得多，其圆形的红色泪珠形状几乎成为一幅抽象画，以此来使它变得陌生。

The cow was Julia Lohmann’s “Waltraud” Cow-Bench. Its headless form was creepy but cute, the gleaming brown leather that it was made from taken off a once-living cow and molded into a cow shape once again. “This cow exists as a reminder. It’s almost like a grounding element,” Antonelli explained. It represents the extraction of resources, animal and environmental, to create rarefied objects. It’s still an object, however; children sometimes climb on it, and the guards must rush in. The Google Maps pin icon was designed by Jens Eilstrup Rasmussen in 2005 and was gifted by Google to MoMA at Antonelli’s request. She decided to print it much larger than it appears on screens, its rounded red teardrop shape almost becoming an abstract painting, as a way of defamiliarizing it.

这三件作品在画廊中构成一幅视觉和概念上都令人震撼的画面。每一件作品在媒介和主题上都独具特色，但安东内利通过将它们并置，却共同营造出一种独特的氛围。或许，这是一种共同的创作手法，即将我们周围的环境从表面彻底颠覆，将熟悉的视角颠倒到令人不适的地步。它们并非相互补充，也并非像Spotify的电台播放列表那样流畅地衔接，而是相互对比和碰撞，展现出每件作品新的面向。画廊角落的这番并置令人叹为观止，令人不禁想起诗人伊西多尔·吕西安·杜卡斯那句标志性的超现实主义诗句：“美得像在解剖台上偶然相遇的缝纫机和雨伞。” 不同物件的碰撞总能带来一种全新的美。“永久收藏和画廊的策展都具有一种非算法的特质，”安东内利说道。

Those three pieces together form a tableau in the gallery, both visually and conceptually striking. Each one is unique in terms of medium and subject, and yet they share a certain tone that Antonelli has drawn out by putting them in proximity. Perhaps it’s a shared approach of turning our environment inside out from its surface, reversing the familiar viewpoint to the point of discomfort. Rather than complementing each other or flowing smoothly together like a Spotify radio playlist, they contrast and clash, bringing out new facets of each work. That corner of the gallery, with its striking juxtaposition, brought to mind the iconic surrealist line by the poet Isidore Lucien Ducasse: “Beautiful as the chance meeting on a dissecting table of a sewing machine and an umbrella.” The collision of different objects always brings out a new kind of beauty. “There’s a non-algorithmic quality to what’s included in the permanent collection and the curation of the galleries,” Antonelli said.

安东内利在画廊里威严地巡视时，注意到一些展品略有瑕疵。在一个房间里，墙上的标签看起来太破旧，而且已经卷边了。在另一个房间里，一台投影仪没有像往常一样投射到展览标题上，导致其字体显示不完整。然后她发现一个画廊缺少保安——他们鉴于这些作品是互动电子游戏，这一点尤为重要。她用手机口述短信给工作人员，一位助理随后赶来解决问题。为了赢得观众的信任，每个细节都必须准确无误。

As she swept authoritatively through the galleries, Antonelli noticed slight flaws in various displays. In one room, the wall labels looked too worn and dog-eared. In another, a projector wasn’t beaming as it should on an exhibition title to complete its lettering. Then she observed that a gallery was missing guards—who were particularly important, given that the pieces were interactive video games. She dictated text messages to her staff on her phone, and an assistant rushed down to address the problems. In order to earn the audience’s trust, every detail must be right.

缓慢的策展过程与互联网特有的无语境、快速和短暂性形成了鲜明对比。“当你想到社交媒体时，它就像白噪音，是死角，”安东内利说道。（这句话恰如其分地描述了整个“过滤世界”，那里没有什么特别突出的东西。）“这时，算法就成了你的对手，”她继续说道。算法推送扰乱了精心策划的并置，使得解读广泛的文化、弄清楚哪些主题将事物联系在一起、哪些方面将它们区分开来变得更加困难。我们没有通过体验获得富有成效的进步，而是陷入了越来越难以区分的泥沼。任何上过网的人都知道，通过推送学习很难。学习，从理解的积累的角度来看，总是发生在平台之外，当你有时间自己去整合东西的时候。即便如此，正如我的艺术家朋友 Hallie Bateman 所说，在互联网上组织材料就像在潮水上涨的沙滩上建造沙堡一样——你精心收集的材料几乎不可避免地会被摧毁，就像 Spotify 的界面变化曾经弄乱了我的界面一样。

The slow process of curation works against the contextlessness, speed, and ephemerality that characterizes the Internet. “When you think of social media, it’s this white noise, it’s dead space,” Antonelli said. (An apt description of Filterworld as a whole, where nothing stands out.) “That’s where the algorithm becomes your antagonist,” she continued. Algorithmic feeds disrupt curated juxtapositions and make it that much harder to interpret the broad swath of culture, to figure out which themes join things together and which aspects set them apart. Instead of a productive progression through experiences, we have an increasingly indistinguishable morass. As anyone who has been on the Internet knows, it is difficult to learn via feeds. Learning, in the sense of the accretion of understanding, always happens off the platform, when you have time to put things together for yourself. Even then, organizing material on the Internet, as my artist friend Hallie Bateman said, is like building sandcastles on the beach as the tide comes in—your careful collections are almost inevitably going to be destroyed, just as Spotify’s interface changes once messed around with mine.

像纽约现代艺术博物馆这样的收藏至关重要，因为它成为我们这个时代的永久记录。市面上有如此多的物品、内容和文物，我们不可避免地需要做出一些选择，决定哪些该消费，哪些该保存。每位策展人都有自己的观点。人类共同构建了我们文化中重要事物的记录，这通常被称为经典。经典可以扩展和变化，不断扩展以包含新的想法，但它的存在是不可避免的。它不仅仅基于受欢迎程度。“有些策展人说，我不再需要经典了，”安东内利说。“但你知道吗？我们无法逃避它，所以我们不妨拥抱它。”在策展人的帮助下，经典不仅包含那些直截了当的美丽或吸引人的事物，也包含那些奇特、令人不安、令人不安或令人震惊的事物。它们不断地促使我们重新思考特定物品或审美体验的意义。

A collection like MoMA’s is vital because it becomes a lasting document of our moment. There are so many objects, pieces of content, and artifacts out there that some choices inevitably have to be made about what to consume and what to preserve. Every curator brings in her own viewpoint. Collectively, humanity builds up a record of what matters in our culture, which is often called a canon. The canon can expand and change, widening to encompass new ideas, but its existence is inevitable. It’s not just based on popularity. “Some curators say, I don’t want a canon anymore,” Antonelli said. “But you know what? We can’t escape it, so we might as well embrace it.” With curators’ help, the canon incorporates things that are strange, disturbing, discomfiting, or shocking as much as the straightforwardly beautiful or appealing. They are constantly urging us to reconsider what a specific object or aesthetic experience means.

安东内利回到办公室，我在纽约现代艺术博物馆闲逛，直到看到法国超现实主义艺术家梅雷特·奥本海姆的回顾展。她生于1913年，卒于1985年。我对奥本海姆知之甚少，只知道她1936年创作的一件著名雕塑，作品由一个茶杯、茶碟和一把勺子组成，表面覆盖着瞪羚皮毛。这件雕塑名为《物件》，有时还会附上“皮毛早餐”的字样。它是超现实主义的标志，皮毛将熟悉的物品变成了既怪诞又引人注目的事物。但她的其他作品则以严谨的时间顺序排列在六个画廊空间，令人惊喜不已，从她年轻时卡通化、焦虑却又有趣的画作，到异教风格的自然神像独立雕塑，不一而足。参观展览令人如沐春风，因为它完整地捕捉了一位艺术家毕生创作的复杂性。尽管奥本海姆生活在不同的时代，但通过艰苦的策展工作，她的存在和视角立即呈现在画廊中。

Antonelli returned to her office, and I wandered around MoMA until I came to a retrospective of the French surrealist artist Meret Oppenheim, who lived from 1913 to 1985. I didn’t know much about Oppenheim save her famous sculpture of a teacup, saucer, and spoon covered in gazelle fur, from 1936, titled Object and sometimes appended with Breakfast in Fur. It’s an icon of surrealism, the fur turning the familiar item into something both bizarre and compelling. But her other work, arrayed in rigorous chronology through half a dozen gallery spaces, was a surprise, from the cartoonish, angsty-yet-funny paintings of her youth to freestanding sculptures of paganesque nature gods. Walking through the exhibition felt like a breath of fresh air, because it captured the entire lifelong creative practice of an individual artist in all its complexity. Though Oppenheim lived in a different time, her presence and perspective were immediately present in the galleries, through the laborious act of curation.

在Instagram上，我或许会在推荐页面上找到奥本海姆的茶杯，或许还能找到她那张光鲜亮丽的黑白肖像。她的作品正是这个平台的成功之作——第一眼就令人惊喜，赏心悦目。这些图片在信息流中闪现时或许会令人振奋，但我却无法从中了解到更多关于奥本海姆的信息，无法了解她与时装公司的合作，也无法了解她在20世纪中期作为女性艺术家罕见的公众曝光度，更不用说看到她早期的涂鸦了。（“没有人会给你自由；你必须自己争取，”她在1974年的一次演讲中说道。）我的观点并非我们必须仅仅通过博物馆展览来理解艺术；而是我们通过算法信息流获得的对文化的理解往往过于狭隘，以至于毫无用处。我们缺乏足够的鼓励或信息去超越它，因为那样并不能为应用程序的广告收入提供素材。人工策展是更广阔、更深层次的视角，最终也是更令人满意的视角。

On Instagram, I might have found Oppenheim’s teacup and perhaps a glamorous black-and-white portrait of the artist on my recommended page. Her art is exactly the kind that succeeds on the platform—visually surprising and pleasing at first glance. The images would be inspiring as they flashed through the feed, but I wouldn’t learn much else about Oppenheim, about her collaborations with fashion houses or her rare public visibility as a woman artist in the mid-twentieth century, much less see her early doodles. (“Nobody will give you freedom; you have to take it,” she said in a 1974 speech.) My point is not that we must gather an understanding of art only through museum exhibitions; rather, it’s that the view we have of culture through algorithmic feeds is often so blinkered as to be useless. We’re not encouraged or informed enough to get beyond it, because that would not provide fodder for the app’s advertising revenue. Human curation is the expanded and deepened view, and ultimately the more satisfying one.

无论我们多么依赖或归因于算法推送，它们的核心功能都是将一条内容放在另一条内容旁边，无论是在 Netflix、Spotify、Facebook 还是 TikTok 上。这些推荐决定了哪些内容应该放在一起，并决定了你的路径。随之而来，这在你的脑海中形成了一个不可避免的叙事。我对于安东内利等人所实践的策展方式的观点是，将一件物品摆放在另一件物品旁边的行为至关重要，应该留给对相关主题有深厚知识或充满热情的人去做——那些在意“接近”意义的人。正如安东内利所说，他们是我们的“值得信赖的向导”。这种排序的做法本身甚至可以成为一种艺术形式。

However much we rely on or ascribe to them, the core function of algorithmic feeds is to put one piece of content next to another one, whether on Netflix, Spotify, Facebook, or TikTok. The recommendations decide what belongs together and dictate what path you will follow, which forms an inevitable narrative in your mind. My argument for the kind of curation that Antonelli and others practice is that the act of putting one thing next to another is an incredibly important one and should be left to people with deep knowledge about or passion for the subject at hand—people who care about the significance of proximity. They are our “trusted guides,” as Antonelli put it. That practice of ordering can even become an art form in itself.

DJ 担任策展人

DJ AS CURATOR

虽然策展是一项重要的行为，是一项基于情感和理智的决定，但这并不意味着它是一种稀缺的体验。我们可以从许多不同的渠道找到策展，而且很容易忽视它，或者认为其中的劳动是理所当然的。在一次格外艰辛的公路旅行中，与自动推送的信息形成鲜明对比的是，策展的重要性让我想起了它。

Though curation is an important act, a decision made on both feeling and intellect, that doesn’t mean it’s a rarefied one. We can find it from many different sources, and it’s easy to overlook it or take for granted the labor that goes into it. I was reminded of the importance of curation, as a contrast to the automated feed, on a particularly arduous road trip.

2022年感恩节后的那个周末，大多数美国人都开始返乡了。我和杰西刚从康涅狄格州回来，开车回华盛顿特区的公寓。时间不太合适，纽约市周围的道路都挤满了人。幸好我们偶然发现了一档广播节目，这让我们在拥堵中找到了喘息的机会。这档节目来自90.7 FM，名为WFUV的电台，始于1947年，总部位于布朗克斯区的福特汉姆大学。那天晚上，一位名叫保罗·卡瓦尔孔特的DJ正在主持他每周一次的Cavalcade节目，从晚上八点到十一点。

The weekend after Thanksgiving in 2022, when most of America gets back on the road to return home, Jess and I were driving to our apartment in Washington, D.C., returning from a visit to Connecticut. It wasn’t ideal timing; the roads surrounding New York City were packed. The saving grace amidst the congestion was a radio show that we stumbled upon by chance. It was on 90.7 FM, a station called WFUV, which began in 1947 and is based at Fordham University in the Bronx. A DJ named Paul Cavalconte was doing his weekly Cavalcade show from eight to eleven p.m. that night.

那时我对卡瓦尔孔特一无所知。但我们在车里听到的是他那流畅、甜美、单调的嗓音，如同台球桌上的绿色毛毡，为我们打开了一条隧道，穿过挡风玻璃，进入他自己的音乐世界。那一周，他的演出以《剩菜剩饭》为主题，反映了感恩节的背景。他创建了一个翻唱和B面的播放列表，这类录音可能无法反映一位音乐家的全部作品，但正因如此，它才更有趣，就像一个深度剪辑的合辑。但随着我们开车，DJ继续他的喋喋不休，我们开始注意到更深层次的结构。卡瓦尔孔特以一种菊花链的方式，一位艺术家翻唱另一位艺术家：Death Cab for Cutie 乐队翻唱了 Cat Power 的《Metal Heart》，然后 Cat Power 翻唱了鲍勃·迪伦的《Stuck Inside of Mobile with the Memphis Blues Again》，接着鲍勃·迪伦翻唱了琼尼·米切尔的《Big Yellow Taxi》。这些翻唱歌曲的并置揭示了创作型歌手领域内的影响联系。每位艺术家都互相欣赏，并发挥出彼此歌曲的特定品质，这些品质也暗示了艺术家自己原创的音乐感知力，比如 Cat Power 用她自己沙哑、讽刺的嗓音突出了迪伦歌声中被低估的旋律和推进力。尤其是在我正在进行算法清理的时候，我非常渴望得到推荐，所以这首翻唱歌曲给我留下了深刻的印象。

I knew nothing about Cavalconte at that point. But what we heard in the car was his smooth, sweetly monotone voice, like the green felt on a pool table, opening a tunnel for us, somewhere beyond the windshield, into his own musical universe. That week, he structured his show around “Leftovers,” reflecting the Thanksgiving context. He had created a playlist of covers and B sides, the kinds of recordings that might not reflect a musician’s entire body of work but were more interesting for it, a compilation of deep cuts. But as we drove and the DJ kept up his patter, we began noticing a deeper structure. Cavalconte made a kind of daisy chain as one artist covered another: the band Death Cab for Cutie covered Cat Power’s “Metal Heart,” then Cat Power covered Bob Dylan’s “Stuck Inside of Mobile with the Memphis Blues Again,” then Bob Dylan covered Joni Mitchell’s “Big Yellow Taxi.” The juxtaposition of covers unveiled links of influence within a realm of singer-songwriters. Each artist was an admirer of the other and brought out specific qualities of their songs that hinted at the artist’s own original musical sensibility as well, like Cat Power highlighting the underrated melody and propulsion of Dylan’s singing with her own smoky, wry voice. Particularly since I was starved for recommendations in the midst of my algorithm cleanse, it made a powerful impression.

就像Paola Antonelli在纽约现代艺术博物馆（MoMA）的展厅里布置艺术品一样，Cavalconte排列歌曲的方式也传达了歌曲的某种特质，构建了一幅更广阔的文化图景。DJ也是策展人——在歌单的背后，我们能感受到他们打造歌单所需的深厚专业知识和敏锐度。Spotify切换到自动播放模式时，我感到很沮丧，因为除了风格或声音的相似性之外，它所组合的曲目之间没有任何联系。相比之下，聆听Cavalconte的表演，我总能感受到其背后的人性智慧，这让它更加引人入胜。就像展览中每件展品上的标签一样，Cavalconte也会定期停下来，分享一些音乐历史或他对特定曲目的个人看法。虽然如今DJ的角色可能不那么引人注目（部分原因是算法的介入），但他们也帮助听众探索新文化，这至关重要。

Just like Paola Antonelli arranging artworks in a MoMA gallery, the way that Cavalconte lined up the songs communicated something about them, building to a broader picture of culture. DJs are curators, too—behind the playlist we could sense the deep expertise and sensitivity that it took to put it together. I get frustrated when Spotify toggles into its automated radio mode because nothing links the tracks that it puts together except genre or a sameness of sound. Listening to Cavalconte’s show, by contrast, I could always feel the human intelligence behind it, which made it infinitely more compelling. Like the labels on each object in an exhibition, Cavalconte also periodically paused to offer some musical history or his personal perspective on a particular set of tracks. While the role might not be as high-profile these days (due in part to the intrusion of algorithms), DJs also help their listeners in that vital process of discovering new culture.

在我写这本书的时候，独立电台DJ一直是我脑海中浮现的、非算法文化传播的理想形式。即使在互联网出现之前，电台也全天候播放着手工挑选的音乐和信息。当通过实际的无线电波访问时，它们也特定于其所在的地理区域（无线电波不可能永远传播），并且实时存在，响应与听众共享的语境——无论是天气、时间还是方言。这样描述，听起来近乎乌托邦，但也很平常：人们通过扬声器与我们对话的奇迹已经变得司空见惯，尽管它曾经像机器分类的数字内容推送一样罕见。但当值得信赖的电台的DJ播放歌曲时，我们仍然倾向于更仔细地聆听，比直接跳到下一个推荐时更专注地听更长时间。

As I’ve written this book, independent radio DJs have stuck out in my mind as an ideal form of non-algorithmic cultural distribution. Even pre-Internet, radio stations beamed out a round-the-clock stream of music and information, all selected by hand. When accessed by actual radio waves, they are also specific to their geographic area (the waves can’t travel forever) and exist in real time, responding to a context that’s shared with the listener—whether that’s the weather, time of day, or a regional dialect. Described this way, it sounds almost utopian, but it’s also mundane: The miracle of people speaking to us through our speakers has become familiar, though it was once as unusual as a digital feed of machine-sorted content. But when a DJ on a trusted radio station plays something, we still tend to listen more closely, pay attention for longer than if we could instantly skip to the next recommendation.

就像艺术展一样，电台会挑战我们对可能性的感知。我高中毕业班的一个晚上就遇到过这种情况。我的方向感很差，经常在家乡蜿蜒曲折的山路上迷路。在从朋友家开车回家的深夜，我会听当地的WPKN电台，这是一个非商业电台，由一群志愿DJ组成，起源于布里奇波特大学。《纽约客》曾半讽刺地称它为“世界上最伟大的电台”；我很幸运能在它的陪伴下长大。2005年左右，周六晚上，一位女士主持了一场爵士和布鲁斯节目；我不知道她的名字，但我对她的声音很熟悉，她的声音又细又沙哑，经常会渐渐消失，还带有独立电台特有的随意插播。我并不是爵士乐或布鲁斯音乐的忠实粉丝，而且在之后的许多年里也依然如此，但 WPKN 的音乐总是很清晰，播放的音乐也比商业电台的音乐有趣得多。

Like an art exhibition, the radio station can challenge our sense of what’s possible. That happened to me one night when I was a senior in high school. My sense of direction is terrible, and I often got lost on the unlit roads that twisted through the hills of my hometown. On late drives back from my friends’ houses, I would listen to the local radio station WPKN, a noncommercial station with a staff of volunteer DJs that originated at the University of Bridgeport. The New Yorker once labeled it, only semi-ironically, “the greatest radio station in the world”; I was lucky to grow up in its range. On Saturday nights around 2005 a woman hosted a jazz and blues show; I didn’t know her name, but I became familiar with her voice, which was thin and raspy and often trailed off into silence, with the characteristic haphazard interruptions of indie radio. I wasn’t a particular fan of jazz or blues, and wouldn’t be for many years afterward, but WPKN always came in clear and played music much more interesting than what was on the commercial stations.

那天晚上，爵士乐曲以一连串强劲的钢琴和弦、鼓点以及立式贝斯的敲击声开场。随后，一段圆号旋律如同流星划过，又如晴空中盘旋的鸟儿。虽然从结构上看，这是一段标准的爵士四重奏，但却与我以往听过的任何音乐都截然不同。我开车时，音乐不断展开，每当我以为它会停下来时，它又会突然加速，时而节奏平静，时而转为不和谐的圆号独奏，偶尔才会出现一些可辨认的旋律。这段音乐持续了十三分钟多，把我一路带回了车道。当歌曲最终结束，收音机里传来女人的声音，说这是约翰·柯川1961年录制的《我最爱的东西》（My Favorite Things），完整版，而不是后来成为热门单曲的那个短得多的电台剪辑版。 （我甚至不知道 1959 年 Rodgers and Hammerstein 创作的原版歌曲。）在这首歌中，Coltrane 演奏了 Miles Davis 拥有的高音萨克斯管这是他前一年买给他的，当时在爵士乐中并不常见。DJ 用简短却信息丰富的音符开始播放下一组歌曲。我坐在驾驶座上，微微一愣。

That night, a jazz track started with a series of pounding piano chords and a drum backbeat alongside percussive thunks from an upright bass. Then a single horn melody traced over the background like a shooting star or a bird whirling against a clear sky. Though structurally it was a standard jazz quartet, it was unlike anything I had ever heard before. The track kept unspooling as I drove, charging along every time I thought it might stop, moving through periods of rhythmic calm and then into discordant horn solos that only dipped occasionally into recognizable melody. It lasted more than thirteen minutes, bringing me all the way back into the driveway. When the song finally subsided, the woman’s voice on the radio explained that it was John Coltrane’s recording of “My Favorite Things” from 1961, the full version rather than the much shorter radio edit that became a hit as a single. (I wasn’t even aware of the original song by Rodgers and Hammerstein from 1959.) On the track, Coltrane played a soprano saxophone that Miles Davis had bought for him the year before, an uncommon instrument for jazz at the time. And with that brief but informative note, the DJ went off into her next set of songs. I sat in the driver’s seat gently stunned.

从那以后，那张录音一直是我最喜欢的音乐作品之一。每个人都有类似的故事，讲述一件艺术品——绘画、音乐、电影——如何让他们惊喜不已，并在瞬间发生难以言喻的变化。但如果没有策展人的帮助，也没有电台等成熟文化渠道的介入，我们就不太可能拥有这样的邂逅。

Ever since, that recording has remained one of my favorite pieces of music. Everyone has a similar story of how a piece of art—painting, music, film—took them by surprise, left them ineffably changed in an instant. But without the help of curators and the context of established cultural channels like radio stations, we’re less likely to have these encounters.

如果 Spotify 上自动播放的是同一首 Coltrane 的歌曲，我可能确实听到了，但我不会像在收音机上听到它那样去听。Spotify 的推荐系统从未向我推荐过这么长的歌曲，也许是因为这种时间投入往往导致我跳过曲目，按照算法的标准，这是一个负面指标。（回想一下流媒体时代歌曲长度的缩短。）我也不会了解到任何关于 Coltrane 或他的影响者的信息。Spotify 会发布艺术家的长篇传记，并开发自己的手工播放列表，但它的界面让人很难找到关于特定专辑的任何信息，除非离开应用程序直接谷歌搜索，甚至连重新发行的原始录制日期都找不到，更不用说哪位音乐家演奏了哪种乐器了。与 CD 光盘盒或黑胶唱片插页不同，Spotify 上没有反映艺术家审美情趣的内页说明。

If the same Coltrane track had autoplayed on Spotify, I might have heard it, in a literal sense, but I wouldn’t have listened to it in the way that I did when it was on the radio. Spotify’s recommender system has never suggested such a long song for me, perhaps because that kind of time commitment too often leads to skipping the track, a negative indicator by the algorithm’s standards. (Recall the streaming era’s shortening song lengths.) Nor would I have learned anything about Coltrane or his influences. Spotify publishes lengthy biographies of artists and develops its own handmade playlists, but its interface makes it difficult to find any information about a particular album without leaving the app to just google it, even the original recording date of a rerelease, much less which musician plays which instrument. Unlike in a CD jewel case or a vinyl gatefold, there are no liner notes that reflect the aesthetic sensibility of the artist.

算法无法取代人类DJ。Spotify或许也意识到了这一点；它在2023年推出了一款AI DJ，可以间歇性地告诉听众算法播放列表中的歌曲名称，但使用它感觉更像是一种侮辱，而不是创新。创造力是必需的。感恩节过后，我和杰西开车回家，听完了保罗·卡瓦尔孔特（Paul Cavalconte）主持的WFUV节目，一口气听了三个小时。当我们在新泽西某处向南行驶时，电台信号中断了，我们就用手机从电台网站上播放了节目。在接下来的几周里，我与卡瓦尔孔特进行了交谈，想弄清楚这个平凡的播放列表奇迹是如何诞生的，以及他如何看待自己在算法推送时代的角色演变。

The algorithm is a bad substitute for a human DJ. Perhaps Spotify is aware; in 2023, it launched an AI DJ that could intermittently tell listeners the names of songs in the algorithmic playlist, but using it felt more insulting than innovative. Creativity is required. As Jess and I drove home from Thanksgiving, we listened to all three hours of Paul Cavalconte’s WFUV show. When we lost the radio signal heading south somewhere in New Jersey, we streamed it from the station’s website over our phones. In the weeks that followed, I struck up a conversation with Cavalconte, wanting to figure out how that ordinary miracle of a playlist had come together, and how he saw his role evolving in the age of algorithmic feeds.

Cavalconte 已经担任 DJ 三十多年了，他看起来就像你想象中的那个人：有点书呆子气，头发顶着一圈光环高额头、大眼睛和灿烂的笑容，传达出极具感染力的热情。他住的地方离 WFUV 的工作室只有 10 分钟车程，就在他长大的家里，后来他把地下室公寓改造成了自己的录音室，周围摆满了成千上万张唱片。他是独生子，父母年纪较大，所以他最早接触音乐是通过他们收藏的古典和爵士乐唱片，上学时，摇滚乐也融入了他的音乐。20 世纪 80 年代初，他就读于福特汉姆大学，在当时完全由学生志愿者运营的广播电台做志愿者。在他成长的时期，在 FM 的鼎盛时期，电台 DJ“是一种文化仲裁者”，他用和我在收音机里听到的一样甜美的声音告诉我，带着一丝纽约口音。

Cavalconte has worked as a DJ for more than thirty years and looks like you’d imagine one: a little nerdy, a halo of hair cresting a high forehead, wide eyes and a wide grin, communicating infectious enthusiasm. He lives only a ten-minute drive away from WFUV’s studio, in the family home he grew up in, where he eventually transformed the basement apartment into his own sound studio, surrounded by shelves of thousands of records. He was an only child, and his parents were on the older side, so his earliest exposure to music came in the form of their classical and jazz record collection, a mixture that was leavened by rock when he went to school. He attended Fordham University during the early 1980s, volunteering at the radio station, which was entirely run by student volunteers at the time. While he was growing up, during the period of peak FM, radio DJs “were a kind of cultural arbiter,” he told me in the same buttery voice, tinged with New York accent, that I heard on the radio.

卡瓦尔孔特的第一份兼职工作是在长岛颇具影响力的前卫摇滚电台WLIR找到的。该电台以允许DJ以更随意的语气播音，并播放更多鲜为人知的专辑曲目而非单曲而闻名。WLIR很早就开始拥抱八十年代流行的新音乐流派，包括后朋克（略带旋律感的七十年代中期朋克的继承者，例如B-52s乐队）和无浪潮（No Wave），后者是源自纽约艺术界的一种更喧闹、更无调性的音乐运动。WLIR播放的音乐至今仍在挑战主流品味。其宗旨是保持文化前沿，并为听众提供一个策展渠道；该电台甚至建立了自己的供应链，以便当天从英国配送最新唱片。在 WLIR 之后，卡瓦尔孔特的职业生涯跨越了纽约的广播电台：WNEW-AM、WNCN、Q104.3、CD101.9、WRXP，所有电台都有自己的格式、个性以及我们现在所说的品牌。

Cavalconte got his first part-time job on the radio at WLIR, an influential progressive rock station based on Long Island. The station was known for allowing DJs to have a more casual tone on the air and play more obscure album tracks, not just radio singles. WLIR was early to embrace new music genres that became popular in the eighties, including post-punk, the slightly more melodic successor to mid-seventies punk, with bands like the B-52s, and No Wave, a noisier, atonal movement emanating from the New York City art scene. The music that the station played still challenges mainstream taste today. Its purpose was to stay on the edge of culture and provide a curatorial channel to listeners; the station even set up its own supply chain for getting same-day deliveries of the newest records from the United Kingdom. After WLIR, Cavalconte’s career crisscrossed New York radio stations: WNEW-AM, WNCN, Q104.3, CD101.9, WRXP, all names that had their own formats, personalities, and what we would now call brands attached.

Cavalconte 也担任古典和爵士乐节目的 DJ，并根据演出需要，灵活应对各种演出形式。2013 年，他重返 WFUV，并于 2015 年创办了 Cavalcade。2017 年，他还接手了 WNYC 的周六晚间和周日下午时段。此前，DJ Jonathan Schwartz 曾负责这两个时段，他以 20 世纪的《美国金曲集》（Great American Songbook）为灵感，带来精彩纷呈的演出，之后因性骚扰丑闻离开该电台。这是一个标志性的时段，实时播放……在新的一周开始之前，纽约市居民准备外出或放松时，营造出纽约市的氛围——这是一项艰巨的策展责任。

Cavalconte DJed classical and jazz shows, too, taking on whatever format the gig required. He returned to WFUV in 2013 and started Cavalcade in 2015. In 2017, he also took over the Saturday night and Sunday afternoon slots on WNYC, which the DJ Jonathan Schwartz had occupied with meandering shows drawing on the twentieth-century Great American Songbook, before he left the station in a harassment scandal. It’s an iconic time slot, broadcasting a real-time atmosphere over New York City as its residents get ready to go out or relax before a new week begins—a hefty curatorial responsibility.

DJ们精心挑选的歌曲之间穿插着玩笑，加上他们各自的嗓音——卡瓦尔孔特说，最好的听起来像“很酷的老师”——“把你拉近了一点，让你稍微涉足反主流文化的更深层领域”。“这是一种灌输；非常具有颠覆性，”他说。说话的时候，卡瓦尔孔特的声音完全进入了广播模式，放慢了语速，吐字清晰。我后颈的汗毛都竖了起来。“在我成长的过程中，”他说，“DJ们的声音都洪亮；听起来真的很性感。我想去他或她带我去的地方——昏暗的灯光和烟灰缸里冒烟的世界。你想象一下电台播音室下班后的氛围，仅仅是某个人的声音就能把你带到这里。”他继续说道，“DJ 是最初的影响者群体的一部分，我们非常羡慕这些人，他们做这些事还能拿到钱——穿得邋遢，喝着免费的啤酒，听着免费的音乐。”

DJs’ selection of specific songs linked by the banter interspersed between, plus their individual voices—the best sounded like “cool teachers,” Cavalconte said—“drew you in just a little, a few toes into the deeper water” of the counterculture. “It was an indoctrination of sorts; it was very subversive,” he said. As he spoke, Cavalconte’s voice slid into full radio mode, slowing down and enunciating each word. The hair stood up on the back of my neck. “When I was growing up,” he said, “the DJs all had these sonorous voices; they were really sexy sounding. I want to go where he or she is taking me—the world of the dimmed light and the smoldering cigarette in the ashtray. You imagine this after-hours ambience of the radio studio, and just the tone of someone’s voice takes you there.” He continued, “The DJ was part of the original community of influencers, these people we were so envious of who got paid to do this—dress like a slob and drink free beers and listen to free music.”

就像博物馆策展人一样，DJ 营造了一种信任的氛围，让消费者能够接受新的文化。“如果一个嗓音好听的男人把你吸引到圈子里，也许会发生一些神奇的事情。我们需要这种陪伴，”卡瓦尔孔特说。“策展现在就是陪伴。”这句话让我想到，算法推送的信息流缺乏陪伴的作用：它们只是把内容呈现给你，然后让你独自面对，直到你失去耐心，决定跳到下一首歌或视频。

Like the museum curator, the DJ cultivates an atmosphere of trust in which the consumer can take in new culture. “If the man with the nice voice has lured you into the circle, maybe something magical will happen. There’s a need for that kind of companionship,” Cavalconte said. “Curation is now companionship.” That statement made me think about how algorithmic feeds are absent as companions: they just surface a piece of content and leave you alone with it until you lose patience and decide to skip to the next song or video.

独立电台的 DJ 表演是一种由他人精心策划的整体沉浸式体验，是一种融入策展人品味和知识的体验。卡瓦尔孔特认为，音乐选择构成了一个“更大的叙事概念”。当然，更多的商业电台必须遵循公司音乐清单和严格的歌曲配额，但在他的每周 WFUV 节目中，卡瓦尔孔特只是探索自己跨越时代和流派的音乐品味——美国标准、民谣、爵士、摇滚、流行、嘻哈——这些品味都是他数十年研究和实践积累下来的。他说：“品味是一个更大的参考框架；它是一系列复杂的价值判断，你一直以来都有意识或无意识地做出这些判断。”推荐算法对策展人来说就像一个对手，就像民间故事里约翰·亨利开着蒸汽机跑着修建铁路隧道一样。这个“东西”——算法——“现在成了你的朋友，你的品味仲裁者，这太恶心了。我不想参与其中；我想要一个真实的人，一个可以投资的人，”卡瓦尔孔特说。

A DJ set on an indie radio station is a holistic immersion that has been planned out for you by another person, an experience informed by the curator’s taste and knowledge. The music choices form a “larger storytelling concept,” according to Cavalconte. More commercial stations have to follow the dictates of corporate music lists and rigid song quotas, of course, but for his weekly WFUV show, Cavalconte simply explores his own music taste across eras and genres—American standards, folk, jazz, rock, pop, hip-hop—built up over decades of research and practice. “Taste is a larger frame of reference; it’s an intricate series of value judgments that you consciously or unconsciously made all along,” he said. The recommendation algorithm is a kind of opponent for the curator, like the steam-powered machine that John Henry raced in the folktale to build a railroad tunnel. The fact that this “thing”—the algorithm—“has now become your friend, your arbiter of taste, it’s gross. I don’t want any part of that; I want a real person, someone to invest in,” Cavalconte said.

不过，数字过滤机器中仍然蕴含着某种隐喻。“我的做法是算法化的，但它源于我自己的思维和经验框架，”卡瓦尔孔特解释道。“我尽量保持自由联想，把自己想象成心理医生，玩那种游戏。如果觉得有点傻，那就放着它。” 在他的感恩节演出中，他加入了几首泰勒·斯威夫特的歌曲。卡瓦尔孔特说，选择这位当代流行歌星，在众多经典音乐家和翻唱歌曲中，堪称“一笔带过”。（你可以试试问问Spotify的算法，是想看看是讽刺还是幽默。）

Still, there’s a certain metaphor in the digital filtering machine. “The way I do it is algorithmically, but it’s generated from within my own mind and frame of my experience,” Cavalconte explained. “I try to be as free association as possible, to put myself on the shrink’s couch and play that game. If it’s something silly, go with it.” During his Thanksgiving set, that took the form of adding in a few Taylor Swift songs. The choice of the contemporary pop star was an “ironic brushstroke” amidst the more classic musicians and cover songs, Cavalconte said. (Try asking the Spotify algorithm for irony or humor.)

人与人之间的推荐是双向的：策展人必须考虑他们所传递内容的价值，而消费者必须保持开放的心态，如果歌曲没有立即吸引人，就放弃跳过它。“你必须让某人接受他们无法掌控的想法。事实上，关键在于你默许了失去这种控制权，”卡瓦尔孔特说道。

Recommendations between people are a two-way exchange: the curator must consider the value of what they are passing on, and the consumer must remain open-minded, giving up the option of skipping the track if it’s not immediately appealing. “You must get somebody to come on board with the idea that they don’t have control. In fact, the whole point is that you acquiesce to the loss of that control,” Cavalconte said.

个性化算法推荐强调熟悉和可识别的内容，将内容定制为最不令人反感的选项，而DJ则致力于突出不熟悉和不寻常的内容。他们无法保证你会喜欢他们播放的内容，但希望你至少会对它感兴趣。这种差异化在整个文化中也很重要。人们很可能对某样东西感兴趣，但并不喜欢它，比如一首难懂的音乐或一幅抽象画。一件艺术品可能会激起你的共鸣，让你感到困惑或不安，但仍然会被它吸引。在Filterworld中，更常见的情况是，你可能喜欢某些东西，但并不觉得它有趣，就像Netflix的《艾米丽在巴黎》一样：它看起来足够令人愉悦，但看完之后，这种体验就像苏打水中冒出的气泡一样，瞬间从你的脑海中消失。（真正感兴趣需要某种质感的存在，而不是完全可以忽略的氛围。）策展是不断进步的，而不是无休止的循环往复。“关键在于让那些原本不知道的人也能获得它，”卡瓦尔孔特说。“你只有在得到它之后才会知道自己想要什么。”

While personalized algorithmic recommendations emphasize the familiar and recognizable, tailoring the content toward the least objectionable options, DJs work to highlight the unfamiliar and unusual. There’s no guarantee that you’ll like what they play, but the hope is that you’ll at least be interested in it. That differentiation is important in culture at large, too. It’s very possible to be interested in something but not like it, in the case of a difficult piece of music or an abstract painting. A piece of art can provoke you and leave you confused or perturbed but still drawn in. Perhaps more commonly in Filterworld, you can also like something but not find it interesting, as in the case of Netflix’s Emily in Paris: it’s pleasant enough to watch, but when it’s over, the experience immediately leaves your mind like the bubbles effervescing in seltzer. (Actual interest requires the presence of some texture, rather than total ignorable ambience.) Curation progresses forward instead of in endless reiterative circles. “It’s a question of making something available to someone who otherwise wouldn’t have known about it,” Cavalconte said. “You don’t know what you want until you’ve got it.”

没有了策展带来的额外摩擦，文化往往会变得越来越千篇一律。卡瓦尔孔特在算法推送时代观察到了音乐的这种效应——他无论在工作还是生活中都对此保持关注。“科技让人们能够更即时地获取音乐，并能够以快速的方式品味音乐，而不是像往常一样坐着听专辑，感受专辑信息以平淡无奇、充满新意的方式逐首展开，”他说道。单曲的重要性已经超越了专辑，专辑也变成了堆积如山的冗余素材，与其说是简洁的陈述，不如说是情绪背景。与黑胶唱片或磁带不同，流媒体的形式不受时间限制；曲目越多，算法推荐和基于流媒体的版税的素材传播范围就越广。

Without the added friction of curation, culture tends to become more and more generic. Cavalconte has observed this effect with music in the era of algorithmic feeds—which he pays attention to both professionally and personally. “Technology has allowed people to have much more instant access to music and to be able to taste-test music in a rapid-fire kind of way, rather than the discipline of sitting and listening to an album and experiencing the prosaic, novelistic way that the message of the album unfolds song by song,” he said. Individual tracks have superseded albums in importance, and albums have become baggy collections of excess material, sprawling more into mood background than concise statements. Unlike a vinyl album or cassette tape, the format of streaming doesn’t impose any time limits; the more tracks there are, the wider the spread of fodder for algorithmic recommendations and royalties based on streams.

以泰勒·斯威夫特的作品为例。她在2020年至2022年间发行了三张新专辑，并发行了两张早期专辑的重新录制版本，这充分体现了Spotify首席执行官丹尼尔·埃克（Daniel Ek）的理念：当代音乐家必须持续发布新音乐。（2020年，埃克表示，艺术家每隔几年发行一张专辑是不够的；他们必须与粉丝保持“持续的互动”。）尽管斯威夫特在2014年曾将自己的音乐从Spotify下架，因为她认为这贬低了她的作品价值，但她最终还是接受了这种源源不断的内容。其中两张原创专辑《Folklore》和《Evermore》是两张节奏轻快的民谣专辑，几乎难以区分。 2022年发行的《Midnights》是一张轻松、反思、合成器音效丰富的专辑，其B面专辑也与之相似。听众听到了更多音乐，但内容却大同小异。 （另一位世界顶级流行歌星 Drake 在这些年里也发布了一系列混音带，包括 2022 年的《Honestly, Nevermind》，在这张混音带中，这位说唱歌手的文字狂热被简化为在环境合成器音效下不断重复的自恋焦虑歌词。）

Take Taylor Swift’s output. She released three new albums from 2020 to 2022 and released two re-recorded versions of her earlier albums as well, fulfilling the Spotify CEO Daniel Ek’s dictate that contemporary musicians must constantly release new music. (In 2020, Ek said that it wasn’t enough for artists to release an album every few years; they have to create a “continuous engagement with their fans.”) Even though, in 2014, Swift had removed her music from Spotify because she felt it devalued her work, she eventually embraced the endless stream of content. Two of those original albums, Folklore and Evermore, are a pair of downbeat folk records that are nearly indistinguishable. Midnights from 2022 was an album of chilled-out, reflective, synth-heavy tracks that came with its own immediate batch of seven similar B sides. Listeners got more music, but it was more of the same. (Drake, another of the world’s biggest pop stars, released a stream of mixtapes over those years as well, including 2022’s Honestly, Nevermind, which saw the rapper’s graphomania reduced into scant, repeating lines of narcissistic angst over ambient synth washes.)

就像 Instagram 的极简主义设计美学一样在算法推送的压力下，室内音乐已然成为一种千篇一律的风格。“一切听起来都像循环，声音单一。节奏比旋律更占主导地位，”卡瓦尔孔特说道，并引用了比莉·艾利什的“卧室流行乐”。艾利什是完全在“过滤世界”时代崭露头角的明星之一。TikTok 视频的简洁性和可快速跳过的特性，将音乐艺术浓缩成几秒长的独立片段，而这正是“声音”吸引用户注意力所需的全部时间。将音乐延续数分钟并非可行之策。

Just like the generic Instagram design aesthetic of minimalist interiors, music has settled into a generic style under the pressure of algorithmic feeds. “Everything sounds like a loop, with one-dimensional sounds. Rhythm is a more dominant characteristic than melody,” Cavalconte said, citing the “bedroom pop” of Billie Eilish, one of the stars who emerged wholly in the era of Filterworld. The brevity and instant skipability of TikTok videos has condensed musical artistry into seconds-long self-contained segments, which is all the time a “sound” has to capture a user’s attention. Unfolding over many minutes is not an option.

“当代音乐中几乎不存在调性变化，因为它不再起作用了，”卡瓦尔孔特说道。他哼唱了几句惠特尼·休斯顿的《我想和某人共舞》，这首歌的高潮部分就是调性变化。这种技巧需要一种音乐叙事来建立旋律的对比，并有清晰的前后衔接。它不能被简化成一段声音片段。“那种悬念的营造——现在什么都没有了。一切都必须在前三十秒内发生，”他说。

“The key change is almost nonexistent in contemporary music, because it’s a hook that doesn’t work anymore,” Cavalconte said. He hummed a few lines of Whitney Houston’s “I Wanna Dance with Somebody,” with its climactic key change. The technique requires a musical narrative that sets up the melodic contrast, with a clear before and after. It’s not reducible to a sound bite. “That building of suspense—now there’s no building of anything. It all has to happen in the first thirty seconds,” he said.

这一观察并非只是DJ的直觉。一位研究人员发现，从20世纪60年代到90年代，Billboard Hot 100榜单中有四分之一的歌曲出现了调性变化，但2010年代只有一首歌曲出现了调性变化。90年代出现了像Biggie的《Juicy》和Tim McGraw的《Something Like That》以及The Streets乐队2004年的概念专辑《A Grand Don't Come for Free》那样漫无目的的叙事歌曲，而流行音乐中普遍的叙事性近年来似乎也逐渐减弱，取而代之的是整体氛围和情绪。歌词避免要求听众过多的注意力。

That observation is not just the DJ’s hunch. One researcher found that a quarter of the songs from the Billboard Hot 100 from the 1960s to 1990s featured a key change, but only one song in the 2010s did. Where the nineties had rambling narrative songs like Biggie’s “Juicy” and Tim McGraw’s “Something Like That,” or the Streets’s 2004 concept album A Grand Don’t Come for Free, storytelling in general in pop music also seems to have lately fallen off in favor of overall vibes and moods. Lyrics avoid requiring too much attention from the listener.

流媒体时代的歌曲通常也很短——格莱姆斯在她2020年专辑《Miss Anthropocene》的豪华版中，发布了几首“算法混音”版本的歌曲，缩短了歌曲的播放时间，使其内容更丰富，更具即时吸引力，更适合算法推送。（这与过去的“电台混音”并无二致。）平均而言，热门歌曲在过去二十年中变得更短，从1995年的4分30秒缩短到2019年的3分42秒，总共缩短了30秒。加州大学洛杉矶分校的数据科学家计算出，2020年在Spotify上发布的歌曲平均长度仅为3分17秒，而且这个长度趋势还在进一步缩短。音乐学家内特·斯隆认为，这种集体缩短是由流媒体的激励机制造成的。服务——例如，Spotify 将 30 秒的收听视为一次“播放”，并根据该标准支付版税。超过 30 秒就没有任何经济效益。

Streaming-era songs are often brief, too—Grimes, for the deluxe version of her 2020 album Miss Anthropocene, released a few “Algorithm Mix” iterations of songs, cutting down their run time, making them denser and more immediately compelling, better for algorithmic feeds. (It’s not dissimilar to the “radio mixes” of the past.) On average, hit songs have gotten shorter in the past two decades, decreasing a total of thirty seconds from 4:30 in 1995 to 3:42 in 2019. Data scientists at UCLA calculated that the average length of a song released in 2020 on Spotify was just 3:17, and that length is trending even shorter. The musicologist Nate Sloan has argued that the collective shortening is caused by the incentives of streaming services—Spotify, for example, counts thirty seconds of listening as a “play” and pays out royalties based on that metric. There’s no financial benefit to going any longer.

这并不是说如今的音乐比过去几十年差，也不是说艺术家们不够努力。（作为一名忠实的氛围音乐爱好者，我个人很欣赏对氛围音乐的接纳。）但显然，文化的默认形式既取决于平台的需求，也取决于艺术家的个人创造力。寻找真人策展人，而不是随波逐流，可以帮助我们打破这种自我强化的循环。策展人的工作不仅是让我们接触新事物，更要帮助我们以一种不同的、或许更真实的方式体验文化。

It’s not that music today is worse than in decades past, or that artists aren’t trying hard enough. (As a committed fan of ambient music, I personally appreciate the embrace of atmospheric qualities.) But it’s clear that the default forms of culture are determined as much by the demands of platforms as the artists’ sense of personal creativity. Seeking out human curators, rather than following the current of the feed, can help us break that self-reinforcing cycle. It’s the curators’ job not only to expose us to new things but to help us experience culture in a different, perhaps more authentic way.

“这就是我的风格——让它变得激进，让它变得有趣，”卡瓦尔孔特说道。他以十五世纪佛罗伦萨画家桑德罗·波提切利为例。波提切利在15世纪80年代创作了《维纳斯的诞生》，画中女神从海中浮现。如今，这幅画作是艺术史上最著名的画作之一，但在当时，它令人震惊，令人感到怪异，甚至连Instagram上都不会有人点赞。品味是一种充满想象力的练习，其目的就是让人感到不适。

“That’s my thing—to make it radical, to make it interesting,” Cavalconte said. He used the example of Sandro Botticelli, the fifteenth-century Florentine painter. Botticelli painted his The Birth of Venus with the goddess emerging from the ocean, in the 1480s. Today it’s one of art history’s most famous paintings, but at the time it was shocking and strange, an image you wouldn’t “like” on Instagram. Taste is an imaginative exercise meant to be uncomfortable.

重振线上策展

REVIVING CURATION ONLINE

策展是一个模拟过程，无法像社交网络信息流那样完全自动化或规模化。最终，它还是要靠人类的审批、选择和安排。但这并不意味着策展不能像在博物馆展览或广播节目中那样存在于网络世界。如今，我们已经见证了2010年代后半期算法信息流在文化领域占据主导地位所带来的扁平化效应，企业家和设计师们正在构建新的数字平台，将策展放在首位，弱化自动推荐的重要性。这些平台的规模远小于Facebook和Spotify等公司，这既是有意为之，也是迫于无奈。

Curation is an analog process that can’t be fully automated or scaled up the way that social network feeds have been. It ultimately comes down to humans approving, selecting, and arranging things. But that’s not to say curation can’t exist online just the way it does in a museum exhibition or on a radio show. Now that we’ve seen the flattening effects of the dominance of algorithmic feeds in culture during the latter half of the 2010s, entrepreneurs and designers are building new digital platforms that put curation first and deemphasize automated recommendations. They are much smaller than the likes of Facebook and Spotify, both intentionally and by necessity.

标准收藏公司 (Criterion Collection) 成立于 1984 年，旨在汇集世界各地重要的当代电影经典，并以各种格式（录像带、CD-ROM 和 DVD）进行授权和出版。标准收藏已成为电影领域最知名、最重要的品牌之一。策展，就像米其林电影指南一样，但可以在家观看。它建立了一个包含一千多部作品的图书馆，收藏了独立热门影片和艺术电影，其中包括一个世纪以来由让·科克托、黑泽明、斯派克·李和阿方索·卡隆等导演创作的影片。几十年来，它从一种媒体形式过渡到另一种，保存了一批原本可能被技术变革冲走的艺术作品。

The Criterion Collection is a company that began in 1984 with the goal of grouping together a canon of important contemporary films from around the world and then licensing and publishing them in various formats: videocassettes, CD-ROMs, and DVDs. Criterion became one of the best-known and most important brands for film curation, a kind of Michelin Guide to cinema but available for consumption at home. It built up a library of over a thousand productions, collecting indie hits and art films, with a century’s worth of material by directors ranging from Jean Cocteau to Akira Kurosawa, Spike Lee, and Alfonso Cuarón. Over the decades, it has moved from one media format to the next, preserving a group of art pieces that otherwise might have been washed away by technological change.

2008 年，Criterion 开始将其服务转移到线上，首先是通过名为 Mubi 的流媒体服务，然后是通过 Hulu，后来是通过特纳经典电影公司创建的另一项订阅流媒体服务 FilmStruck。它也可以在 Kanopy 上使用，用户可以通过图书馆和其他机构访问该服务。但在 2018 年，FilmStruck 被其所有者华纳媒体关闭。《纽约时报》称这一消息“令人震惊”，并在该服务消失之前列出了在该服务上观看的最佳影片列表。这促使 Criterion 推出了自己的数字流媒体服务，称为 Criterion Channel。如今，只要有互联网，就可以在任何地点使用 Criterion，它是 Netflix 的超级精选版本。其图书馆配备了丰富的观影指南、导演历史访谈、影评人视频文章以及及时的电影精选。这些定制推荐和背景材料使其有别于 Netflix 的算法主页。

In 2008, Criterion began moving its offerings online, first through a streaming service called Mubi and then through Hulu and later FilmStruck, another subscription streaming service that was created by Turner Classic Movies. It was also available on Kanopy, a service users could access through libraries and other institutions. But in 2018, FilmStruck was shut down by its owners under Warner Media. The New York Times described the news as “devastating,” and lists popped up of the best titles to watch on the service before it disappeared. It prompted Criterion to launch its own digital streaming service, called the Criterion Channel. Today, Criterion is available anywhere with Internet access, a hypercurated version of Netflix. The library comes with a rich set of viewing guides, historical interviews with directors, video essays from critics, and timely selections of films. Those bespoke recommendations and contextual material set it apart from Netflix’s algorithmic home page.

我采访了佩内洛普·巴特利特 (Penelope Bartlett)，她曾担任 Criterion Channel 的节目总监，任期至 2022 年。巴特利特在电影节和制片人岗位工作后，于 2016 年开始担任 Criterion 的节目策划员——负责挑选影片，也算是另一种形式的策展人。她说，这份工作“很像艺术影院的节目策划员：挑选主题影片，将影片组合成有趣且引人入胜的套装”。这可以是特定导演或特定演员的回顾展，回顾他们的职业生涯，也可以是安排特别精彩的双片放映。

I spoke with Penelope Bartlett, who was the Criterion Channel’s director of programming until 2022. Bartlett started as a Criterion programmer—people who choose which films to feature, another kind of curator—in 2016, after working at film festivals and as a producer. The job is “really akin to what a programmer does at an art-house theater: selecting thematic programming, putting films together in interesting, appealing packages,” she said. That could take the form of retrospectives of specific directors or specific actors, progressing through their careers, or setting up particularly good double features.

“这确实是标准收藏的工作人员精心策划的，而不是某种算法，”巴特利特说。即使有算法推荐的帮助，网上内容的广度也会导致一种决策瘫痪。“人们常常被网站上琳琅满目的选择弄得不知所措。”“流媒体领域。我们花了好几个小时琢磨看什么，”巴特利特继续说道。Criterion 的节目制作“只是想帮助人们以既刺激又易于理解的方式欣赏和发现电影，一个晚上就能看完。”

“It’s truly thoughtfully curated by the staff of the Criterion Collection versus some kind of algorithm,” Bartlett said. Even with the assistance of algorithmic recommendations, the breadth of content available online can induce a kind of decision paralysis. “People are often overwhelmed by the multitude of options in the streaming space. We spend hours figuring out what to watch,” Bartlett continued. Criterion’s programming is “just trying to help people enjoy and discover movies in exciting ways that also feel manageable, something that you can get through in an evening.”

巴特利特将自己的角色描述为“牵着人们的手”。“有时候，人们会被这些拍了三十部电影的电影人吓到。你真的不知道从哪里开始；最终你一部都没看，因为你不确定切入点在哪里，”她说。基于观看次数最多的电影的算法推荐可能并非最佳选择。Criterion 的目标远不止表面的推荐：“它不仅仅是我应该看什么，而是我为什么要看，以及我还能用它看什么。” Criterion 充当着一种内容认证的印章，其评判标准并非由观众数量或销量决定，而是由其策展团队纯粹的艺术品质决定。

Bartlett described her role as an attempt to “hold people’s hands.” “Sometimes people are a little bit intimidated by these filmmakers who have made thirty movies. You don’t really know where to start; you end up not watching any of them because you’re not sure what the entry point is,” she said. An algorithmic recommendation based on which movie was watched the most times might not be the best choice. The purpose of Criterion goes beyond superficial recommendation: “It’s not just what should I watch, but why should I watch it, what else could I watch with it.” Criterion acts as a kind of content seal of approval, not dictated by audience numbers or sales but by sheer artistic quality, as determined by its staff curators.

标准影业帮助我发现了自己独特的电影感性。香港导演王家卫已成为我最喜欢的电影人之一，因为他那光彩夺目的视觉美学，​​以及他叙事中那种缓慢、怀旧、浪漫的特质。我第一次接触他2000年的电影《花样年华》是在我十几岁的时候，在一家本地百视达的外国电影书架上，DVD盒的封面上印着标准影业的标志。这部电影对20世纪中叶香港双性婚姻的刻画，以及缺失的爱情纽带（更不用说剧中人物外卖面条的饮食），至今仍萦绕在我的心头。多亏了这家公司，以及一位匿名但备受赞赏的百视达员工，我才得以接触到王家卫的作品。从《花样年华》开始，我又看了他早期的电影《重庆森林》，这是一部关于沮丧警察的黑色喜剧，还有《2046》，这是一部更加晦涩、带有科幻色彩的《花样年华》续集。

Criterion helped me discover my own film sensibility. The Hong Kong director Wong Kar-wai has become one of my favorite filmmakers, for his lambent visual aesthetics and the slow, nostalgic, romantic quality of his storytelling. I first encountered his 2000 film In the Mood for Love as a teenager in a local Blockbuster, on the foreign-film shelf, in a DVD case with the Criterion logo on the cover. The film’s aching portrait of dual marriages in mid-century Hong Kong and the missed connections of love (not to mention the characters’ diets of takeout noodle soup) have stuck with me ever since. It was thanks to the company, and an anonymous but much appreciated Blockbuster employee, that I was introduced to Wong’s work at all. From In the Mood for Love, I made my way to his earlier film Chungking Express, a comedic noir about a bummed policeman, and 2046, a more obscure, science-fiction-inflected sequel to In the Mood for Love.

如果没有偶然发现百视达，我可能完全错过了黄家驹的作品，因为我人生的成长期，无法去电影节或保留剧目的艺术影院——但现在所有这些电影都可以在网上观看了。流媒体平台 Criterion Channel 将互联网的即时、不受地域限制的可访问性与图书馆或博物馆深度负责的策展相结合——两者兼得，又免去了无休止的干扰。无限的信息流。这是我们对互联网如何运作的愿景，如果我们决定它应该如何运作的话。

Without the serendipitous Blockbuster discovery, I might have totally missed Wong’s work in that formative period of my life, since I lacked access to film festivals or repertory art-house theaters—but all these films are now available online. The streaming Criterion Channel combines the instant, nongeographic accessibility of the Internet with the deep, responsible curation of a library or museum—the best of both worlds, without the relentless distraction of unbounded feeds. It’s a vision of how the Internet could work if we decided it should.

网上文化付费

PAYING FOR CULTURE ONLINE

迈向更加精细化的互联网的另一个步骤是更加仔细地思考驱动我们所使用平台的商业模式。互联网上有一句格言，可能起源于2010年MetaFilter论坛上的一条评论：“如果你不付费，你就不是顾客；你只是被出售的产品。” 当数字平台可以免费使用并通过广告赚钱时，内容就沦为一种吸引注意力的方式。然而，当你直接为内容本身付费时，内容在经济上更具可持续性，往往会投入更多资源，这对创作者和消费者都有利。在Netflix和Spotify等流媒体服务上，用户需要支付订阅费才能访问内容，但这笔资金分散在平台上的所有内容上——而且随着新订阅量的增长放缓，这两家公司都越来越多地转向广告获取收入。

Another step toward a more curated Internet is to think more carefully about the business models that drive the platforms we use. There’s an axiom about the Internet that might have originated in a comment on the forum MetaFilter in 2010: “If you are not paying for it, you’re not the customer; you’re the product being sold.” When digital platforms are free to use and make money through advertising, content is reduced to a way of attracting attention. When you are paying directly for the content itself, however, the content is more economically sustainable and tends to have more resources invested into it, which is better for both creators and consumers. On streaming services like Netflix and Spotify, users are paying subscription fees for access to content, but that funding is spread across all the content on the platform—and both businesses are increasingly turning to advertising for revenue as their growth in new subscriptions has slowed.

规模较小的流媒体服务正在提供一个替代的生态系统，用户可以订阅一组特定的精选内容，从而更好地支持其创作者。Criterion 是流媒体视频的一个选择，但也有一些专注于英国电视剧、惊悚电影或动漫的流媒体服务。Idagio 是一家专注于古典音乐的流媒体服务，于 2015 年推出，其创始人 Till Janczukowicz 告诉我，它是一种“公平贸易流媒体”——就像咖啡公司承诺对咖啡农进行非剥削性交易一样。Spotify 根据超过 30 秒的收听次数向艺术家和唱片公司支付费用，而 Idagio 提供的录音比普通流行歌曲长得多，它根据用户收听特定唱片公司的时间百分比支付费用，精确到秒。例如，如果用户将 30% 的时间用于收听德意志留声机 (Deutsche Grammophon) 的录音，那么这家拥有百年历史的德国古典音乐厂牌将获得该用户 30% 的订阅版税。根据经验，内容创作者从平台获得的报酬越直接、比例越高，平台的可持续性就越强。

Smaller-scale streaming services are providing an alternative ecosystem, where you can subscribe to a specific set of curated content and better support its creators. Criterion is one option for streaming video, but others exist that focus on British television, thriller movies, or anime. Idagio, a streaming service solely for classical music, launched in 2015 as a form of “fair-trade streaming,” as its founder, Till Janczukowicz, told me—like coffee companies that promise a nonexploitative deal for the coffee farmers. While Spotify pays out to artists and labels based on number of listens that last over thirty seconds, Idagio, which hosts much longer recordings than the usual pop song, pays out based on the percentage of their time a user listens to a particular label, down to the second. If a user spends 30 percent of their time listening to recordings from Deutsche Grammophon, for example, then the century-old German classical music label receives 30 percent of that user’s subscription royalties. As a rule of thumb, the more directly and proportionally content creators get paid by a platform, the more sustainable it is.

互联网过去几十年的发展很大程度上建立在尽可能免费的内容访问之上。在20世纪90年代和21世纪初，网站似乎只是企业微不足道的额外支出，而非主流产品。随后，随着谷歌搜索和AdSense的兴起，广告成为互联网的主要商业模式。流量越大，收入也就越高，因此限制访问显得毫无意义。随着社交网络的兴起，Facebook、Twitter和YouTube等公司通过算法推送吸引用户注意力并销售自有广告，从而迫使内容创作者和出版商与系统博弈，从而吸收了更多的广告收入。直到近年来，我们才开始意识到，销售注意力可能比销售内容本身和避免推送中介更难，也更不可持续。直到2011年3月，任何人都可以在线阅读《纽约时报》。当时，它实施严格的付费墙并向读者收取数字内容费用的决定被视为冒险且非正统的。但它如今已成为全球最成功的新闻公司之一，拥有超过900万数字订阅用户，因为它很早就拥抱了付费墙。虽然直接付费似乎会让网络文化更加商品化，但实际上，文化本身就应该是产品，就像古典音乐在Idagio上，而不是你的关注点。

The past decades of the Internet were largely premised on content being as free to access as possible. In the 1990s and early 2000s, websites seemed like an insignificant side expense for businesses, not mainstream products. Then, with the rise of Google Search and AdSense, advertising became the primary business model of the Internet. Since more traffic meant more money, it didn’t make sense to restrict access. As social networks emerged, companies like Facebook, Twitter, and YouTube absorbed more of the advertising revenue by capturing users’ attention with algorithmic feeds and selling their own ads, forcing creators and publishers to game the system. Only in recent years have we begun to realize that selling attention can be harder and less sustainable than selling the content itself and avoiding the mediation of feeds. Up until March 2011, anyone could read The New York Times online. Its decision to implement a strict paywall and charge readers for its digital content was seen as risky and unorthodox. But it is now one of the most successful journalism companies on the planet, with more than nine million digital subscribers, because it embraced the paywall so early. While it may seem as though culture online is more commodified when it is paid for directly, in reality, culture should be the product, as classical music is on Idagio, not your attention.

Janczukowicz 从传统的古典音乐界转向数字技术，因为他发现他所热爱的文化在互联网时代正在落后。Janczukowicz 是一位德国音乐经理人，他五岁开始弹钢琴，但逐渐转向古典音乐创作、组织工作坊、录制唱片，并代理著名音乐家和指挥家。他曾与中国钢琴家郎朗、日本指挥家小泽征尔和芬兰指挥家尤卡-佩卡·萨拉斯特等人合作。在关注这些音乐人的职业生涯时，他注意到他们的作品并不像预期的那样容易获得。“如果它们在数字空间中无法被检索到，它们就会失去意义，”Janczukowicz 告诉我。他总结了 Idagio 的目标：“我们并非致力于打造流媒体服务。我们的主要驱动力，以及背后的使命，是如何利用技术来传承一种文化？” （“维护文化”正是 Filterworld 在加速追逐最低标准的过程中未能做到的。）

Janczukowicz moved from the traditional classical music world into digital technology because he saw the culture that he loved falling behind in the Internet era. A German impresario, Janczukowicz began playing piano at the age of five but gradually moved into writing about classical music, organizing workshops, producing recordings, and managing famous musicians and conductors. He worked with the likes of the Chinese pianist Lang Lang, the Japanese conductor Seiji Ozawa, and the Finnish conductor Jukka-Pekka Saraste. As he shepherded their careers, he noticed that their work wasn’t as available as it could be. “If they aren’t retrievable in digital space, they are losing relevance,” Janczukowicz told me. He summarized the goal of Idagio: “We are not here to build a streaming service. The main driver, the mission behind it, is how can we use technology to maintain a culture?” (“Maintaining culture” is precisely what Filterworld fails to do, in its accelerated race to the lowest common denominator.)

Janczukowicz 认为，虽然 Spotify 的算法不可避免地会将用户引向流行音乐，但 Idagio 旨在支持更小众的艺术创作。该服务“内容齐全”，这意味着 Idagio 已安排了超过 200 万首曲目的授权协议，尽可能地囊括了有史以来制作的所有古典音乐唱片，包括 2500 个管弦乐队和 6000 指挥家的唱片。每月只需 10 美元，该服务就能提供一个完美的数字图书馆——一个可听的维基百科。你可以在那里找到任何你想找的古典音乐唱片，然后根据专辑的任何细节进行进一步探索。当然，对于像古典音乐这样有限的流派来说，这比对于更广泛的流派更容易实现。但与许多科技公司不同，Idagio 并不不惜一切代价追求规模；它不需要无限扩张并涵盖所有流派。正如 Criterion 的 Bartlett 告诉我的那样：“我们不需要达到那些大型流媒体服务的水平。如果我们拥有忠诚的、专注的观众，我们就应该能够维持下去，达到我们赖以生存的水平。” 除了技术之外，资本主义不惜一切代价追求增长的心态也是 Filterworld 文化扁平化的根本原因。

While Spotify’s algorithm inevitably shepherds the user toward popular pop music, Janczukowicz argued, Idagio is built to support more niche artistic creations. The service is “content complete,” meaning that Idagio has arranged licensing deals for more than two million tracks, getting as close as possible to hosting every recording of classical music ever produced, including twenty-five hundred orchestras and six thousand conductors. For ten dollars a month, the service provides something like a perfect digital library—a listenable Wikipedia. You can find any classical recording you’re looking for there, and then explore further based on any detail of the album. Of course, this is easier to accomplish for a limited genre like classical than a wider one. But unlike many tech companies, Idagio doesn’t pursue scale at all costs; it doesn’t need to expand infinitely and address every genre. As Bartlett from Criterion told me, “We don’t need to be at the level of those huge streaming services. If we have a loyal, dedicated audience, we should be able to sustain ourselves at the level that we need to be able to continue.” Beyond the technology, the capitalist growth-at-all-costs mindset is also fundamentally to blame for the flattening of culture in Filterworld.

同时满足所有类型的用户是一种刻意为之的决定，但往往以牺牲可用性为代价。就像 Facebook 扩张到涵盖所有形式的在线内容一样，流媒体服务也经常试图为所有消费者提供所有内容。追求能够吸引最多人（或至少避免冒犯最多人）的东西会导致同质化。而这种同质化不可避免地会以主流群体的模式塑造：白人、顺性别者、异性恋。通过一个旨在同时适用于数十亿人的算法模型来开发个性化身份非常困难。相比之下，构建专注于更特定主题的小型消费社区可以带来更深层次的参与感，无论是与内容还是用户之间。小规模的可持续性仍然算作成功。而这正是我们在互联网优先考虑无摩擦便利性和同时向尽可能多的人传播时所忽略的。

Catering to every kind of user at once is an intentional decision that often comes at the cost of usability. In the same way that Facebook expanded to encompass every form of content online, streaming services have often attempted to provide all things for all consumers. The path of chasing something that will appeal to, or at least avoid offending, the highest number of people leads to homogeneity. And that homogeneity is inevitably cast in the mold of dominant groups: white, cisgender, heterosexual. It’s hard to develop an individualized identity through an algorithmic mold meant to apply to billions of people at once. By contrast, building smaller communities of consumption devoted to more specific subjects can lead to a much deeper sense of engagement, both with the content and among the users. Sustainability at a small scale still counts as success. That is something we’ve missed as the Internet has prioritized frictionless convenience and broadcasting to as many people as possible at once.

Janczukowicz 说道：“我不相信这种按下按钮就能立即听到音乐的便利。”Spotify 的界面。虽然你仍然可以随时在 Idagio 上播放任何你想听的音乐，但“我还对音乐背景感兴趣”。例如，听众可能会在听完柴可夫斯基的音乐会后，想找更多这位俄罗斯作曲家的交响曲，或者寻找同一个乐团演奏贝多芬交响曲。Spotify 几乎无法实现这种精细的搜索。但 Idagio 的界面采用了 Dieter Rams 设计的清晰几何图形，并强调文本而非图像，让这一切变得轻而易举。你可以按作曲家、表演者、作品或时间顺序浏览，每位表演者的身份都会被识别，现场录音和录音室录音的标签也有所不同。只需点击按钮即可获得纸质 CD 小册子的 PDF 版本。这是一个启示，它解答了 Paola Antonelli 对数字平台的批评，即无差别的“死角”。你不会意识到大多数平台为了算法推荐而缺失了多少信息和导航，直到它被恢复。

“I don’t believe in this isolated convenience of pushing a button and you get music immediately,” Janczukowicz said, evoking the interface of Spotify. While you can still stream anything you want at any time on Idagio, “I’m also interested in context.” A listener might leave a Tchaikovsky concert and want to find more symphonies by the Russian composer or look for the same orchestra playing a Beethoven symphony, for example. Spotify makes that level of granularity next to impossible. But Idagio’s interface, which has the crisp geometry of a Dieter Rams design and an emphasis on text rather than images, makes it effortless. You can browse by composer, performer, composition, or chronology, with every performer identified and live recordings labeled differently from studio sessions. PDFs of the printed CD booklets are available at the click of a button. It’s a revelation, the solution to Paola Antonelli’s critique of digital platforms as undifferentiated “dead space.” You don’t realize how much information and navigation is missing from most platforms, gone in favor of algorithmic recommendations, until it’s restored.

订阅 Idagio 时，我很高兴找到了我最喜欢的法国作曲家埃里克·萨蒂 1888 年创作的Gymnopédie No. 1。它那轻柔、蜿蜒、稀疏的钢琴独奏旋律，就像在雨天漫无目的地散步一样，在 TikTok 上非常受欢迎，它为超过 150,000 个视频提供了配乐，内容涵盖从水母镜头到坠入爱河的个人故事。我立即在 Idagio 上找到了几十个不同版本的演绎，由世界各地的音乐家在过去的半个世纪里演奏过。其中包括 Jeroen van Veen 演奏的特别缓慢而洪亮的版本，Noriko Ogawa 1890 年的钢琴演奏，以及 Francis Poulenc 1951 年更为欢快的演绎，当时这首曲子并不像今天这样出名。萨蒂最初只是指出应该“缓慢而悲伤地”演奏；每张录音都让我对音乐有了不同的视角，以及它随着时间的推移是如何被重新诠释的。在Idagio上，平台的结构会根据其特定内容进行调整，从而提升文化体验。相比之下，在TikTok上，最常使用的Gymnopédie录音中甚至没有Satie的名字；而在Spotify上，专辑目录的差异使得追踪表演者和日期变得非常麻烦。

When I subscribed to Idagio, I was delighted to find my favorite composition by the French composer Erik Satie, Gymnopédie No. 1 from 1888. Its gentle, meandering, sparse solo piano melody, like taking an aimless walk on a rainy day, has become very popular on TikTok, where it provides the soundtrack to more than 150,000 videos that range from footage of jellyfish to personal stories about falling in love. I immediately pulled up a list of dozens of different performances of the piece on Idagio, played by musicians across the world and over the course of the last half century. They included a particularly slow and sonorous version from Jeroen van Veen, a Noriko Ogawa performance on a piano from 1890, and a more upbeat performance by Francis Poulenc from 1951, when the piece wasn’t nearly as well known as it is today. Satie originally noted only that it should be played “slow and sorrowful”; each recording gave me a different view of the music and how it had been reinterpreted over time. On Idagio, the structure of the platform is adapted to its specific content, improving the cultural experience. On TikTok, by contrast, Satie isn’t even named on the Gymnopédie recording that is used most often, and on Spotify, variations in album cataloging make keeping track of performers and dates a mess.

随着我更多地使用 Idagio，我发现自己以一种我从未有过的有机方式漫步在古典音乐的走廊里无需这项服务，我就能追随像赵在赫这样的钢琴家，或者深入肖邦的夜曲，聆听众多不同的演奏家和作品。我对古典音乐知之甚少，但这款产品让我的体验变得愉悦，甚至比收听古典音乐电台更轻松。无需借助算法推送，在线托管海量文化内容完全有可能；毕竟，文化本身就提供了一种可供遵循的算法，因为每位艺术家都会影响和启发其他艺术家，参考并构建历史。保罗·卡瓦尔孔特正是基于这种联系构建了他的电台播放列表。我想到了扬祖科维奇关于“维护文化”必要性的描述。音乐不仅需要可以在线收听；它必须以连贯的方式呈现，以一种超越被动消费的方式进行教育。这同样适用于任何其他文化形式。如果你喜欢某样东西，为什么不去了解更多，深入挖掘呢？

As I used Idagio more, I found myself wandering through the corridors of classical music in an organic way that I never could have without the service, following a particular pianist like Jae-hyuck Cho or delving into Chopin’s nocturnes, listening to many different performers and pieces. I have very little knowledge of classical music, yet the product made the experience enjoyable and less intimidating even than listening to a classical radio station. It is very much possible to host enormous bodies of culture online without the help of algorithmic feeds; after all, culture itself offers a kind of algorithm to follow, as each artist influences and inspires others, referencing and building on history. Those are the kind of connections from which Paul Cavalconte built his radio playlists. I thought about Janczukowicz’s description of the need to “maintain a culture.” It’s not just that the music needs to be available to be listened to online; it’s that it must be presented in a coherent fashion, in a way that allows for education beyond passive consumption. The same applies to any other cultural form. If you enjoy something, why not learn more about it and dive deeper?

在与策展人的交流中，我发现大型数字平台完全缺乏一种关怀和照顾的氛围。这些平台将所有文化视为内容，不加区分地大量投放，并鼓励消费者始终停留在表面。对于YouTube来说，一个视频与下一个视频毫无二致；重要的是你是否会点击它，从而接触到更多广告。但这种美好的关怀，既关乎观众，也关乎当下的文化——确保以恰当的方式呈现一件艺术品、一张专辑或一部电影——使得与文化的相遇更加美好。对于艺术家来说，这同样有益，因为他们的努力往往被理解和欣赏，而这恰恰是艺术的目标。

In my conversations with curators, I found a tone of caring and caretaking that is missing entirely from massive digital platforms, which treat all culture like content to be funneled indiscriminately at high volume and which encourage consumers to stay constantly on the surface. For YouTube, one video is the same as the next; all that matters is whether you’re likely to click on it so that you’re exposed to more advertising. But that beautiful sense of care, both for the viewer and for the culture at hand—being sure to present an artwork, album, or film in the right way—makes the encounter with culture so much better. It is better for the artist just as much in terms of being understood and appreciated for what they tried to accomplish, which is so often the goal of art.

我们转向艺术寻求联系，但算法信息流却给我们带来纯粹的消费。真正的联系需要放慢速度，甚至脱离信息流的掌控。你不可能一边阅读CD小册子，一边保持算法流状态。

We turn to art to seek connection, yet algorithmic feeds give us pure consumption. Truly connecting requires slowing down too much, to the point of falling out of the feed’s grip. You can’t stay in an algorithmic flow state while reading a CD booklet.

返回动态

RETURNING TO THE FEEDS

算法清理的最初几周很艰难，但后来我适应了，时间也过得很快。我一点也不想再回去了。回到数字世界的那个角落。但戒断了三个月后，我决定回归，很大程度上是因为新闻周期。作为一名报道科技的记者，我必须了解网络上发生的事情，而2022年末，随着埃隆·马斯克对推特的收购，社交媒体本身成了当时的首要话题。

The initial weeks of my algorithm cleanse were difficult, but then I adapted and time passed quickly. I had no burning desire to return to that corner of the digital realm. But after three months of abstinence, I decided to go back, in large part due to the news cycle. As a journalist covering technology, I had to know what was going on online, and the primary subject of the moment in late 2022 had become social media itself, with Elon Musk’s fraught acquisition of Twitter.

我也怀念对朋友们生活的周全了解。虽然我很庆幸不用被朋友们在Instagram上发布的科莫湖度假照片轰炸，但我却没能看到他们推荐的书籍、他们做的美味佳肴，或是他们可爱的宠物快照，而所有这些，我的算法都会确保呈现。我勉强重新登录，将应用程序恢复到手机上。我的拇指很快就恢复了之前的习惯。

I also missed my ambient awareness of friends’ lives. While I appreciated not being bombarded with Instagram photos of their vacations on Lake Como, I did not get to see their book recommendations, the nice meals they cooked, or their cute pet snapshots, all of which my algorithm made sure to deliver. I grudgingly logged back in and restored the apps to my phone. My thumbs quickly relearned their old patterns.

但我惊讶地发现，即使在我离开社交媒体的短暂时间里，我的大脑化学成分也发生了变化。逃离算法推送就像变成了素食主义者，然后看到一块多汁的牛排——曾经可能吸引人的东西现在却令人反感。我的消费速度慢了下来，在选择阅读、收听或观看的内容时也更加谨慎。当我回来时，我发现推送内容太快、太混乱，太不符合时间顺序了。

But I was surprised to find that something had changed in my brain chemistry, even during my relatively short absence from social media. Escaping algorithmic feeds was like becoming a vegetarian and then seeing a juicy steak—what once might have been appealing was now off-putting. My pace of consumption had slowed, and I was much more deliberate in selecting what to read, listen to, or watch. When I returned, the feeds felt too fast and chaotic, too far from chronological.

随着我的耐受力再次提升，我逐渐习惯了它们。但那种厌恶感依然萦绕在我的心头。这段时间的休息让我更加深刻地理解，推特上的那些戏剧性场面与线下的日常生活几乎没有任何关联。（我从一开始就需要吸取这个教训，这说明我当时是多么沉浸其中。）当我不再花太多时间在那些点赞和粉丝都有意义的平台上时，那种游戏化地追求帖子点赞就失去了意义。我问杰西，在戒除毒瘾期间，我的行为或态度是否有所改变；她说，我最初的暴躁情绪逐渐消退，整体上变得更加平静，“对网上发生的事情不再那么缺乏安全感了”。（不过，她确实抱怨过我无法在Instagram上发布她的好照片。）我发现自己更加注意朋友们在聚会时看手机，正是因为我的手机上没有那么多让我感兴趣的东西。

Gradually I got used to them again as my tolerance ramped up once more. But the feeling of aversion stayed with me. The time off had made me understand in a more visceral way that the drama of the Twitter feed had very little to do with day-to-day life offline. (That I needed to learn the lesson in the first place speaks to how immersed I was.) The gamified pursuit of likes on my posts lost meaning when I wasn’t spending so much time on platforms where the likes and followers were meaningful. I asked Jess if my behavior or attitude had been any different while on the cleanse; she said that my initial crankiness gave way to being calmer overall and “less insecure about what was happening online.” (She did complain that I wasn’t able to post any good photos of her on Instagram, however.) I found myself noticing more when friends looked at their phones while hanging out, precisely because there wasn’t as much to interest me on mine.

就像生产低焦油香烟的烟草公司一样，算法推送本身就制造了它们所宣传要解决的问题。我们不需要每天接触成千上万条按我们觉得最吸引人的内容排列的数字更新。即使我们没有完全脱离网络，按时间顺序推送信息，并鼓励我们少发帖而不是多发帖，或许对我们更有利，也对文化更有利。

Like tobacco companies manufacturing low-tar cigarettes, the algorithmic feeds create the problems they are marketed as solving. We don’t need to be exposed to thousands of digital updates a day arranged by what we might find most compelling. Even if we don’t log off entirely, chronological feeds and incentives to post less instead of more might be better for us, and better for culture.

我没有沉湎于事物的表面，而是利用没有算法的时间去探究自己感兴趣的事物，弄清楚自己在漫无目的的在线时光中究竟在寻找什么。专注于少数几个文化创作者让我更满意。当我收到Spotify Wrapped 2022年的回顾——这份年度品味总结——时，它告诉我，我是比尔·埃文斯（Bill Evans）听众中排名前0.01%的人，这位富有创新精神的爵士钢琴家的职业生涯在20世纪60年代达到顶峰。虽然这个数字看起来很极端，甚至令人尴尬，但我明白我为什么会得到这个数字。我不断创作的配乐是比尔·埃文斯三重奏1961年在先锋村（Village Vanguard）的音乐会，这是一张完整的录音，涵盖了三张唱片。我对这张专辑的每一个细节都了如指掌，从《Gloria's Step》第一次录制时不可避免的磁带故障，到音乐会结尾两次连续录制的《Jade Visions》。 《Jade Visions》是一首哀歌般的歌曲，旋律简洁，充满冥想，由节奏感十足的低音和弦支撑，由三重奏贝斯手斯科特·拉法罗创作。他于1961年晚些时候在一场车祸中丧生，年仅25岁。我越听这张专辑，就越有感触，仿佛它是我收藏中唯一的一张CD。即使任何算法推荐都会让我被其他音乐家或专辑所吸引，我还是坚持听下去。

Rather than floating on the surface of things, I had used my algorithm-free time to investigate what I found interesting and figure out what I was looking for in my aimless time online. I was more satisfied by concentrating on just a few cultural creators. When I got my Spotify Wrapped recap of 2022, that summary of my taste for the year, it informed me that I was in the top 0.01 percent of listeners to Bill Evans, the innovative jazz pianist whose career peaked in the 1960s. Though the number seemed extreme, almost embarrassing, I understood why I had gotten it. My constant writing soundtrack had been the Bill Evans Trio’s 1961 concert at the Village Vanguard, a complete recording that stretched over three discs. I got to know every second of the album, from the unavoidable tape glitch on the first take of “Gloria’s Step” to the two back-to-back takes of “Jade Visions” that end the concert. “Jade Visions,” an elegiac, meditatively simple track buoyed by rhythmic bass chords, was composed by the trio’s bassist, Scott LaFaro, who died in a car accident later in 1961, at the age of twenty-five. I got more from the album the more I listened to it, as if it were the only CD in my collection. I kept listening long after any algorithmic recommendation would have swept me away to other musicians or albums.

我发现，对抗泛泛之谈的方法是寻求具体的，无论你被什么吸引。你无需成为有资质或专业的专家才能成为鉴赏家。你无需将你的观点作为影响者货币化，才能使其合法化。算法承诺取代你的品味，并将其外包给你，就像一个机械肢体，但形成你自己的品味只需要思考、意图和用心。策展是人类行为的一个自然方面：就像我们选择吃什么食物或搭配什么颜色的衣服一样，我们会自然而然地形成对哪些文化吸引我们、哪些不吸引我们的看法。

I found that the way to fight the generic is to seek the specific, whatever you are drawn toward. You don’t need to be a credentialed or professionalized expert to be a connoisseur. You don’t need to monetize your opinion as an influencer for it to be legitimate. The algorithm promises to supplant your taste and outsource it for you, like a robotic limb, but all it takes to form your own taste is thought, intention, and care. Curation is a natural facet of human behavior: Just as we select which food to eat or which colors go together in an outfit, we organically form opinions about which pieces of culture appeal to us and which don’t.

Spotify 电台或 TikTok“为你推荐”等技术带来的轻松消费，有其特定的时间和地点。鼓励，但我担心其根本的被动性正在贬低整个文化创新的价值，并降低我们对艺术的享受。文化建立在个性化推荐而非自动化推荐的基础上，因为我们会分享、解读和回应我们喜爱的事物。这种人为的推荐过程可以像给朋友发送一个他们可能喜欢的东西的链接，并附上几句解释他们为什么喜欢它的话一样简单——开启一段关于文化对你们双方意味着什么的对话。

There’s a time and a place for the kind of lean-back consumption that technology like Spotify radio or the TikTok “For You” feed encourages, but I worry that its fundamental passivity is devaluing cultural innovation as a whole, as well as degrading our enjoyment of art. Culture is built on personal recommendations, not automated ones, as we share, interpret, and respond to the things that we love. That human process of recommendation could be as easy as sending a friend a link to something they might like, along with a few words about why they might like it—starting a conversation about what culture means to both of you.

我的建议

MY RECOMMENDATION

以一篇个人推荐来结束本章似乎很合适。我最初是通过算法推送发现的，但现在它已经成为我品味和自身的重要组成部分。这是日本音乐家佐藤宏 1982 年的一张名为《觉醒》的专辑。在我看来，它不仅是有史以来最好的专辑之一，还完美地展现了本书记录的 Filterworld 的各种力量。我第一次接触佐藤是通过 YouTube 的推荐算法。我偶然发现了专辑中的一首名为《说再见》的曲目，但在 YouTube 上，它被一个名为 Boogie80 的频道上传了一个重新发行的版本，名为《这个男孩》。它几乎没有其他信息，但封面上那张令人回味的图片——一个晒黑的男人在海里游泳——以及用模块化合成器演奏的轻快的开头琶音立刻吸引了我。

It seems fitting to end this chapter with a personal recommendation, something that I initially discovered through algorithmic feeds but has come to feel like a fundamental part of my taste and of myself. It’s an album from 1982 called Awakening, by the Japanese musician Hiroshi Sato, and beyond being—in my opinion—one of the best albums ever made, it’s a perfect demonstration of the various forces of Filterworld that this book has documented. My first exposure to Sato came from YouTube’s recommendation algorithm. I stumbled upon a track from the album called “Say Goodbye,” but on YouTube it was a reissued version titled “This Boy,” uploaded by a channel called Boogie80. It came with little other information, but the evocative cover image of a tanned man swimming in the ocean and the airy opening arpeggio played on a modular synthesizer immediately grabbed me.

我多次打开 YouTube 链接，惊叹于歌曲完美的流行简洁和凄婉的英文歌词，哀悼分手——“希望你不会孤单”——之后才开始进一步探索佐藤的作品。这首歌似乎存在于真空中，尽管它的点击量超过两百万。当我开始听完整的《觉醒》时， “说再见”仍然是唯一一首让我印象深刻的歌曲。然而，音乐的某些方面让我着迷。它不完全是软摇滚，也不是我联想到的美国八十年代音乐的那种夸张的声音。在专辑中的几首歌中，比如“只有一段爱情”，加拿大裔澳大利亚情歌歌手温迪·马修斯作为嘉宾演唱，她高亢的嗓音听起来就像霓虹灯照亮的夜总会的缩影。

I pulled up the YouTube link dozens of times and marveled at the song’s perfect pop concision and plangent English-language lyrics, mourning a breakup—“hope you won’t be lonely”—before I bothered to explore Sato’s work any further. The song seemed to exist in a vacuum, even though it had over two million views. When I started listening to the full Awakening, “Say Goodbye” was still the only song that stuck out to me for a while. Something about the musical palette kept me intrigued, however. It wasn’t quite soft rock, nor was it the over-the-top sound I associated with American eighties music. On a handful of the album’s tracks like “Only a Love Affair,” the Canadian Australian torch singer Wendy Matthews guest starred with a soaring voice that sounds like the epitome of a neon-lit nightclub.

佐藤的专辑成了我做饭时必听的音乐——疫情期间，它每天都会完整播放——我和杰西经常在厨房里跟着它跳舞，沉浸于艺术的魅力，仿佛置身于另一个世界。《觉醒》收录了一首名为《忧郁的音乐》（Blue and Moody Music）的歌曲的两个版本，歌词描绘了一幅深夜练琴到清晨，在钢琴中寻求慰藉的场景。第一个版本是佐藤独自一人演奏，节奏轻快，键盘华丽的演奏。很容易想象这位音乐家坐在三角钢琴旁，俯瞰着灯火通明的城市天际线。但第二个版本将这首歌提升到了一个超凡脱俗的境界，强劲的合成器颤音近乎热带风情，背景中不断重复的电吉他，以及马修斯充满歌剧风格的嗓音，如同划过夜空的流星，佐藤粗犷的嗓音则与之形成鲜明对比。录音的节奏只是加速到结尾；它并非停止，而是在盛放中逐渐消散，仿佛要永远持续下去。这真是天才之作，一首独一无二的音乐，既完全属于那个时代，又完全永恒。然而，Spotify 的播放量显示，《忧郁忧郁音乐（温迪版）》是这张专辑中不太受欢迎的曲目之一，只有 28 万次收听，而 YouTube 热门歌曲《说再见》的收听量超过 300 万次。

Sato’s album became the kind of music that I put on when I cook dinner—it got a daily play in full during the pandemic—and Jess and I reliably danced to it in the kitchen, indulging in a brief moment of art’s ability to transport you somewhere else entirely. Awakening includes two versions of a song called “Blue and Moody Music,” whose lyrics sketch a scene of practicing the piano late at night into the morning and taking solace in the instrument. The first recording is Sato alone, at a downbeat pace with keyboard flourishes. It’s easy to imagine the musician at a grand piano overlooking some illuminated city skyline. But the second version takes the song to a transcendental level, with driving synth trills that are almost tropical, constant riffing electric guitar in the background, and Matthews in full operatic mode like a star shooting across that night sky, with Sato’s rougher voice singing backup in a complementary contrast. The recording only accelerates toward its ending; rather than stopping, it fades out in full bloom, as if to continue forever. It’s genius, a singular piece of music that’s both completely of its time and completely timeless. Yet Spotify’s play count tells me that “Blue and Moody Music (Wendy’s Version)” is one of the less popular tracks on the album, with 280,000 listens to more than three million for “Say Goodbye,” the YouTube hit.

如此杰作是如何来到我身边的？我开始研究它的起源，以满足自己的痴迷。佐藤是日本著名的钢琴家、制作人和词曲作者，享年数十年，直至2012年去世。但他对美国听众的曝光源于算法。2010年代中期，YouTube的推荐开始关注一种名为“City Pop”的日本音乐流派，这是一种模糊的音乐运动，兴起于20世纪70年代末80年代初。它始于像Happy End这样的乐队，这支在东京颇具影响力的乐队率先用日语歌词创作摇滚和迷幻民谣。Happy End乐队没能持续多久，但其成员包括细野晴臣，他与佐藤等许多音乐家合作，并在更前卫的乐队Yellow Magic Orchestra中尝试使用合成器。 （细野后来为第一家无印良品商店创作了氛围电子音乐配乐，其美学风格与 Filterworld 非常契合。）

How had such a masterpiece made its way to me? I began to research its origins as a way of sating my own obsession. Sato was a famous pianist, producer, and songwriter in Japan for decades, until his death in 2012. But his exposure to American listeners was algorithmic in origin. In the mid-2010s, YouTube’s recommendations began fixating on a Japanese genre called “City Pop,” a nebulous musical movement that emerged in the late 1970s and early ’80s. It began with bands like Happy End, an influential outfit in Tokyo that was the first to make rock and psychedelic-folk music with Japanese lyrics. Happy End didn’t last long, but its members included Haruomi Hosono, who collaborated with many other musicians like Sato and experimented with synthesizers in the more avant-garde band Yellow Magic Orchestra. (Hosono later created an ambient electronic-music soundtrack for the first Muji store, aesthetics that fit very much within Filterworld.)

佐藤、细野等人开始将冲浪摇滚和游艇摇滚融合在一起。来自美国像海滩男孩这样的乐队所创造的摇滚之声，本身就带有科技先锋的感性。1977 年，评论家藤野清一（被学者 Moritz Sommet 在 2020 年的一篇权威论文中引用）将“城市音乐”描述为具有“都市感”的音乐。不过，这位评论家也指出，这个词“并没有什么特别深刻的含义”，而是“看似理解，实则不然”。换句话说，某种模糊性或许是其本质的一部分。它闪烁其词，你投射给它的一切，它都会反射回去。山下达郎 1978 年的专辑《太平洋》（细野参与制作）就是一个很好的例子：它带有明显的夏威夷风情，有松键吉他和海浪的氛围音效样本，几乎可以说是媚俗的。但这张专辑原本充满热带风情，却以完美无瑕的风格收录，最后一首歌却呈现出完全合成的、不和谐的机械音景，仿佛这座岛屿实际上是一个机械化的科幻反乌托邦世界。光滑的表面有时也具有欺骗性。

Sato, Hosono, and others began merging the surf-rock and yacht-rock sounds coming out of the United States from bands like the Beach Boys with their own tech-forward sensibility. In 1977, the critic Tōno Kiyokazu (quoted in an authoritative 2020 paper by the scholar Moritz Sommet) described “City Music” as music with an “urban feeling.” The critic also noted, though, that the term “doesn’t hold any particularly deep meaning.” It is “something that looks like you understand it, but you don’t.” In other words, a certain ambiguity might be part of its nature. It is evasive, reflecting back whatever you project at it. Tatsuro Yamashita’s 1978 album Pacific, which Hosono worked on, is an indicative example: It has blatant Hawaiian influences, with slack-key guitar and atmospheric sound samples of ocean waves. It’s almost kitschy. But the final track on the otherwise immaculately tropical album is a fully synthesized, discordant, robotic soundscape, as if the island was actually a mechanized sci-fi dystopia. Slick surfaces can be deceiving.

城市流行乐还受到另一项技术创新的影响：1979 年索尼随身听的发明。这款设备的诞生源于索尼前高管井深大 (Masaru Ibuka) 的想法，他希望能够在国际航班上收听长篇古典音乐录音。于是，他请求公司为他制作一款便携式音乐聆听设备。工程师们通过改造便携式录音机实现了这一目标。井深大非常喜欢这款设备，于是将其交给了公司董事长盛田昭夫 (Akio Morita)，后者决定生产这款设备。（这是一个直觉决定，而非基于市场预测，因为这款设备本身是前所未有的。）索尼售出了数十万台随身听。突然之间，无论听众身在何处，无论他们选择哪种音乐，音乐都能包围他们。与算法推送一样，随身听是一种极具个性化的形式。 1984年，日本音乐学学者细川周平在《大众音乐》杂志上发表了一篇题为《随身听效应》的文章，文中写道：“听众似乎切断了与他真正生活的外部世界的听觉联系：追求他‘个人’聆听区域的完美。” 有了随身听，物理现实就会顺应听众的心情，就像推荐系统会根据用户的愿望调整数字空间一样。

City Pop was also influenced by another technological innovation: the invention, in 1979, of the Sony Walkman. The device was created because Masaru Ibuka, a former executive of Sony, wanted to be able to listen to long classical music recordings on international flights. So he asked the company to make him a portable music-listening device, which the engineers accomplished by modifying a portable tape recorder. Ibuka liked it so much that he passed it on to the company’s chairman, Akio Morita, who decided to manufacture it. (It was a gut decision, not based on market predictions, because the device itself was wholly unprecedented.) Sony sold hundreds of thousands of them. Suddenly, music could envelop the listener wherever they went, whichever music they chose. Like algorithmic feeds, the Walkman was a dramatic form of personalization. In a 1984 article for the journal Popular Music titled “The Walkman Effect,” the Japanese musicology scholar Shuhei Hosokawa wrote that the “listener seems to cut the auditory contact with the outer world where he really lives: seeking the perfection of his ‘individual’ zone of listening.” With the Walkman, physical reality conformed to the listener’s mood, the same way recommendations bend digital spaces toward users’ desires.

这款设备创造了一种对音乐的需求，使其成为一种移动的、几乎可以忽略的生活配乐。汽车是经济繁荣的另一个好处，它日本中产阶级工薪阶层日益增多，过去常常在周末开车离开东京去海滩冲浪，这为氛围音乐的聆听提供了另一种机会。City Pop 是适合散步闲逛、购物或坐火车时听的音乐。一些 City Pop 音乐家将歌曲作为商业配乐出售或创作，参与到资本主义消费主义的爆炸式增长中，用叮当作响的吉他和管弦乐的号角，赋予其阳光浪漫的氛围。

The device created a need for music as a mobile, semi-ignorable soundtrack to life. Cars, another perk of the boom economy, which the growing population of middle-class Japanese salarymen used to take weekend drives out of Tokyo to beaches to surf, provided another opportunity for ambient listening. City Pop was music for walking and wandering around, going shopping, sitting on a train. Some City Pop musicians sold or wrote songs as commercial soundtracks, participating in the explosion of capitalist consumerism, lending it their sunny, romantic air with jangly guitar and orchestral horns.

八十年代最初的城市流行乐风潮在几年内就消退了，但这种音乐类型仍然以尘封的黑胶唱片的形式流传至今。后来，到了21世纪初，一些在唱片店翻找旧专辑的日本DJ重新发现了这一音乐类型，并开始将其重新带回大众视野。（考虑到数字文件及其存储平台的相对短暂性，未来这种复兴可能会更加困难。）这种复兴通过小众论坛和博客传播到海外，激励西方DJ们计划前往东京购买自己的唱片。随后，城市流行乐登陆YouTube，成为全球主流音乐。

The original eighties fad for City Pop faded within a few years, but the genre stuck around in physical form as dusty vinyl records. Then, in the 2000s, Japanese DJs digging through crates of old albums in record stores rediscovered the genre and began bringing it back into rotation. (Such a revival might be much more difficult in the future, given the relative ephemerality of digital files and the platforms that host them.) The revival spread abroad, through niche forums and blogs, inspiring Western DJs to plan trips to Tokyo to buy their own records. Then City Pop hit YouTube and went mainstream around the world.

不知何故，像佐藤这样的音乐为推荐算法提供了理想的解决方案。其中一首歌尤其火爆：竹内玛莉亚1984年创作的《Plastic Love》。这首歌旋律轻快，R&B节奏轻快，键盘合成器轻柔，旋律优美。这是一首纯粹的流行歌曲。竹内玛莉亚清澈的嗓音盖过了节奏部分，唱出了从心碎中恢复过来的感受——“爱情只是一场游戏/我只需要尽情享受”——其间穿插着英文歌词：“我知道这就是塑料爱情。” 2017年，一个名为“Plastic Lover”的账号在YouTube上上传了该歌曲的版本，获得了超过6300万次观看，只有通过算法推送的推广，这样一首鲜为人知的歌曲才能获得如此高的观看量。事实上，这位匿名上传者也是通过推送才第一次接触到这首歌的。正如他们在 2021 年接受Pitchfork撰稿人 Cat Zhang 采访时所说：“人们告诉我，这首歌总是出现在他们的推荐中。我也遇到过这种情况——我不是第一个上传这首歌的人。起初我对它并不感兴趣，但它却一直出现在我的推荐中。” 算法本身才是这首歌受欢迎的原因。

For some reason, music like Sato’s provided an ideal solution for the recommendation algorithm. One track took off in particular: Mariya Takeuchi’s bouncy 1984 “Plastic Love,” an earworm with a shuffling R&B backbeat and soft synth keyboards. It’s a pure pop confection. Takeuchi’s clear voice rises above the rhythm section, singing about recovering from heartbreak—“Love is just a game / All I need is to have fun with it”—interspersed with lyrics in English: “I know that’s plastic love.” A version uploaded on YouTube in 2017 by an account named Plastic Lover netted over sixty-three million views, an amount only possible for such an obscure track through the algorithmic feed’s promotions. In fact, the anonymous uploader was first exposed to it by the feed as well. As they told the Pitchfork writer Cat Zhang in a 2021 interview, “People tell me they get the song in their recommendations all the time. That happened to me too—I wasn’t the first uploader of the song. At first I wasn’t really interested in it, but it kept haunting me in my recommendations.” The algorithm itself was responsible for its popularity.

关于 City Pop 在网上的火爆，人们提出了各种各样的解释。一些作家认为，这归因于其受欢迎程度YouTube 上充斥着低保真冷门音乐，这种轻松的电子音乐为学习或工作提供了氛围音乐。这些音乐流吸引了数百万听众，随后他们可能会被推送到另一个合成器、低调的中速音乐来源。张在 2021 年的另一项Pitchfork调查中发现：“算法会简单地将听众从‘低保真节拍’视频引导到《Plastic Love》。” 过去几年里，City Pop 流派成了该平台本身的代名词。在 Rate Your Music 网站上，一位用户甚至将其称为“日本 YouTube 推荐核心”，收集了观看次数超过十万的 City Pop YouTube 视频。就《Plastic Love》而言，这首歌的销量可能得益于 YouTube 视频中歌手那张充满幸福感的黑白照片——一张充满自由和无忧无虑幸福感的动人照片，竹内直子咧嘴一笑，眼睛睁得大大的，动作让他的眼睛略显模糊。“歌曲和照片之间有着完美的契合，”摄影师艾伦·莱文森告诉张。就像 Instagram 上白色背景下的色彩鲜艳的陶瓷花瓶一样，该图像经过优化，可以作为 YouTube 推荐缩略图进行传输。

Various explanations have been proposed for City Pop’s runaway success online. Some writers have connected it to the popularity of lo-fi chill music streams on YouTube, the relaxed electronica that provides ambient music for studying or working. Those streams attract millions of listeners, who might then be pushed to another source of synthy, unobtrusive, mid-tempo music. “The algorithm will simply route listeners from ‘lo-fi beats’ videos to ‘Plastic Love,’ ” Zhang found in another 2021 Pitchfork investigation. In the last few years, the City Pop genre became synonymous with the platform itself. On the website Rate Your Music, one user even identified it as “Japanese YouTube Recommendations Core,” collecting City Pop YouTube videos with over one hundred thousand views. In the case of “Plastic Love,” the track may have been helped along by the beatific black-and-white photo of its singer that covered the YouTube video—an evocative image of total freedom and carefree happiness, Takeuchi grinning, eyes wide, blurred slightly by motion. “There’s a perfect kismet between the song and the photo,” the photographer, Alan Levenson, told Zhang. Like a brightly colored ceramic vase against a white backdrop on Instagram, the image was optimized to transmit itself as a YouTube recommendation thumbnail.

除了算法推广之外，这些歌曲本身也确实很棒，出自才华横溢、创造力巅峰的音乐家之手，尽管它们问世时未必就一炮打响。（与2000年代的潮人信条相反，流行并不意味着不好，默默无闻也并不意味着好。）这些歌曲并非偶然流行；它们是为了吸引听众而创作的，尽管最终吸引到的听众规模是无法预测的。

Alongside any algorithmic promotion, it’s also true that these are simply good songs, written by talented musicians at a peak of their creativity, though they were not necessarily hits when they came out. (Contrary to the 2000s hipster credo, just because something is popular doesn’t mean it’s bad, and obscurity doesn’t make something de facto good.) The songs are not good by accident; they were written to reach an audience, though the scale of the audience they ultimately ended up reaching could not have been predicted.

或许就像 Damon Krukowski 的 Galaxie 500 歌曲“Strange”（独立于乐队其他作品在 Spotify 上走红）一样，City Pop 拥有最正常、最悦耳、最普通的流行音乐，让人欲罢不能。这种音乐类型鲜为人知，但却是国际平台上用户都能接触到的普通音乐风格。City Pop 的听觉美学部分源于其日本创作者，部分源于其西方灵感；部分源于其怀旧之情，体现在 2010 年代对 80 年代音乐的感性解读中；部分源于其对当时新兴音乐科技的拥抱，例如合成器和电子鼓机，体现在对未来主义的探索中。City Pop 拥有一种如同垃圾食品般密集的诱人元素。高亢的歌声、厚重的合成器音效、R&B 乐器和鼓点的惯性相结合，让人无法抗拒，而其未经充分审查的日本起源故事使其具有足够的异国情调，以至于西方观众在网上并不熟悉。

Perhaps like Damon Krukowski’s Galaxie 500 track “Strange,” which went viral on Spotify separate from the rest of the band’s work, City Pop possessed the most normal, agreeable, average sound of pop that no one could click away from. The genre was obscure, and yet it was an average of musical styles that could be found accessible by users across an international platform. City Pop’s aural aesthetic was part Eastern, in its Japanese creators; part Western, in its inspirations; part nostalgic, in its eighties sensibility as seen from the 2010s; part futuristic, in its embrace of music technology that was new at the time, like synths and electronic drum machines. City Pop possesses a kind of junk-food density of appealing ingredients. The combination of soaring vocals, heavy synthesizer washes, R&B instruments, and drumbeat inertia is irresistible, and its underexamined Japanese origin story made it exotic enough to be unfamiliar to Western audiences online.

这个名字本身可能有所帮助：“城市流行乐”的含义模糊，可以与任何城市、任何地方联系起来，就像我通过Airbnb注意到的AirSpace美学一样，只不过它指的是音乐。安迪·库什（Andy Cush）在《Spin》杂志上撰文，评论了在YouTube上流行的类似日本氛围音乐：“如果音乐不够好，可能会感觉像个陷阱。” 当然，这就是Filterworld的陷阱。算法推送塑造了一种足够引人注目的文化形式，但它却被彻底地脱离了语境，传播得如此广泛，以至于变得空洞无意义，提供了很多没有内容的美学。

The name itself likely helped: “City Pop” is vague, identifiable with any city, anywhere, like the AirSpace aesthetic I noticed through Airbnb, but for music. Writing in Spin, Andy Cush observed of similar Japanese ambient recordings that became popular on YouTube, “If the music weren’t so good, it might feel like a trap.” Of course, this is the trap of Filterworld. Algorithmic feeds mold a form of culture that is compelling enough and yet decontextualized so fully and spread so widely that it becomes empty and meaningless, offering so many aesthetics without content.

从某种程度上来说，City Pop 的音乐已经经历了这种清空的过程。2015 年《日本时报》的一篇文章指出，这一音乐类型是“一个简化的独立音乐流行词，用来引发成熟、时尚和怀旧的感觉”。随着 City Pop 逐渐饱和其潜在受众，并逐渐让他们感到厌倦，算法推荐将不得不为数字消费的磨坊寻找新的素材。我最近偶然发现了“印尼 City Pop”的提及，它指的是同一时期亚洲其他地区的音乐。一段 2020 年 12 月上传的 YouTube 视频收录了几十首旋律甜美、合成器驱动的软摇滚曲目，观看次数已接近 200 万次。视频标题为“雅加达夜游——80 年代印尼创意流行/City Pop/Jazz Megamix”——一个经过搜索算法优化的词汇组合。在音乐下方播放的视频是一组无限循环的夜间动漫城市景观，这些景观与日本的关系比与印尼的关系更密切。

In some ways, that emptying process has already happened for City Pop. A 2015 article in the Japan Times observed that the genre was “a simplified indie buzzword used to induce feelings of sophistication, fashionableness and nostalgia.” And as City Pop saturates its potential audiences, gradually boring them in turn, algorithmic recommendations will have to find new grist for the mill of digital consumption. I recently came across references to “Indonesian City Pop,” which was music from the same era elsewhere in Asia. A YouTube video uploaded in December 2020 collecting a few dozen tracks of sweetly melodic synth-driven soft rock has almost two million views. Its title is “Jakarta Night Drive—80s Indonesian Pop Kreatif/City Pop/Jazz Megamix”—a search-algorithm-optimized hash of words. The video that plays underneath the music is a set of infinitely looping nighttime anime cityscapes that have more to do with Japan than Indonesia.

在“雅加达夜路”视频中，一种特定的文化被简化为一种虚无缥缈的情绪，被尽可能快速地、尽可能广泛地在线采用、复制和传播，从而吸引肤浅的互动，进而为创作者和平台带来广告收入。视频下的一条评论总结道：“是算法把我从韩国带到这里来的。”在信息流推荐的引导下，全球用户集体聚焦于一组特定的文化主题，就像无数帝王蝶本能地迁徙到墨西哥某片特定的冷杉林一样。

In the “Jakarta Night Drive” video, a specific culture has been reduced to a vaporous mood, to be adopted, replicated, and distributed online as fast and as far as possible, attracting shallow engagement that in turn drives advertising revenue for the creator and the platform. One comment on the video sums it up: “Algorithm brought me here from Korea.” Guided along by the feed’s recommendations, a global population of users collectively converge on a particular set of cultural themes like so many monarch butterflies instinctively migrating to a particular grove of fir trees in Mexico.

这些主题或许是人类文化的根本共性，是那些我们情不自禁地喜爱的事物：短小的歌曲、一致的节奏、戏剧性的视觉清晰度、鲜艳的色彩、妙语连珠的幽默以及富有争议的论点。但更有可能的是，Filterworld 的关注点是由当今文化所流经的数字平台的结构所决定的，而少数审美模式的乏味普遍性正是平台全球化和垄断的结果。Filterworld 包含一个根本的、不可避免的现实：在人类历史上，从未有如此多的人体验过相同的事物，相同的内容通过信息流即时传播到我们每个人的屏幕上。所有后果都源于这一事实。

These themes may be the fundamental commonalities of human culture, the things we can’t help but love: short songs, consistent backbeats, dramatic visual clarity, bright colors, punch-line humor, and controversial arguments. But more likely, the fixations of Filterworld are dictated by the structures of the digital platforms culture now flows through, and the boring ubiquity of a few aesthetic modes is the consequence of the platforms’ globalization and monopolization. Filterworld consists of one fundamental, unavoidable reality: never in human history have so many people experienced the same things, the same pieces of content disseminated instantly through the feeds, to our individual screens. Every consequence flows from that fact.

在YouTube上看到那段“雅加达夜路”的视频，无所谓好坏。事实上，它是一件很酷的艺术品，让我得以一窥国际文化遗产的某个角落，如果没有这个平台，我永远也看不到它。但更重要的是，作为用户，在看完视频、在音乐萦绕脑海之后，你会怎么做。你可以让它随波逐流，相信推荐算法有一天会把它带回来，无论何时，你都会再次欣赏它。或者，你可以找到这首合辑的DJ，并给他们一些小费，感谢他们的文化策划。或者，你可以购买其中一首歌曲或专辑的数字拷贝。或者，你可以研究印尼流行音乐的历史，自己琢磨一下，在独裁者苏哈托统治下，国际资本主义在印尼日益增长的影响力是如何演变的。相比仅仅让视频内容充斥着审美氛围，后两者中的任何一种，都更有利于文化的延续和发展，也更有利于你作为消费者的满足感。

It is neither good nor bad to encounter that “Jakarta Night Drive” video on YouTube. In fact, it’s quite a cool artifact, a glimpse into a corner of the international heritage of culture that I never would have seen without the platform. But what matters more is what you as a user do after you see the video, after the music sticks in your head. You could let it drift by and trust that the recommendation algorithm might bring it back someday, and you’ll enjoy it once more whenever it does. Or you could identify the DJ of the compilation and pay them a tip for their cultural curation. Or you could buy a digital copy of one of the songs or albums that are included. Or you could research the history of Indonesian pop music and chart for yourself how it followed the growing influence of international capitalism in the country under the dictator Suharto. Any one of these latter choices would be better for the continued survival and strength of culture, and your satisfaction as a consumer, than allowing the feed to wash over you with the aura of the aesthetic alone.

为了抵制“过滤世界”，我们必须再次成为自己的策展人，为我们所消费的东西负责。重获这种掌控并不难。你做出个人选择，开始有意识地探索自己的文化兔子洞，这会引导你走向新的方向，做出更加独立的决定。随着时间的推移，这些决定会逐渐形成一种品味，最终形成一种自我意识。

To resist Filterworld, we must become our own curators once more and take responsibility for what we’re consuming. Regaining that control isn’t so hard. You make a personal choice and begin to intentionally seek out your own cultural rabbit hole, which leads you in new directions, to yet more independent decisions. They compound over time into a sense of taste, and ultimately into a sense of self.